Add route info for clusterroles and clusterrolebindings

build(deps-dev): bump electron from 10.1.1 to 10.1.2 in /app

Permissions check & optional metrics-server

Bumps [electron](https://github.com/electron/electron) from 10.1.1 to 10.1.2.
- [Release notes](https://github.com/electron/electron/releases)
- [Changelog](https://github.com/electron/electron/blob/master/docs/breaking-changes.md)
- [Commits](https://github.com/electron/electron/compare/v10.1.1...v10.1.2

)
Signed-off-by: dependabot[bot] <support@github.com>

frontend: Fix joining rules details in the Role details view

We didn't have route info set for clusterroles and clusterrolebindings
because of which when in a clusterrole or clusterrolebinding view we try
to go back we are redirected to cluster chooser page.

This patch adds the info for both routes.

docker: Include a plugins dir in the image if it exists

Fix class names

Otherwise plugins cannot use this class and it will not be possible to
create such objects either.

The name should be CustomResourceDefinition instead of just CRD.

Makefile: Remove --no-cache Docker flag

backend: Add new rules to golangci and update the code to honor them

Some of those rules details are undefined, so we need to default to
empty lists.

If the user cannot list pods or events, then we need to make the UI
aware of that.

The mentioned views were a bit complex and had some problems like:
* Not showing the loading state;
* Not refreshing the list if an error occurred (old results would still
  be shown);
* Not showing any errors.

This patch refactors those views and fixes the mentioned issues.

If the user doesn't have the permissions to list some objects, then we
should reflect that in the UI. This patch does that by replacing the
use of KubeObject.useApiList by KubeObject.useList and using its error
to create an error message when needed.

The RoleBinding and Role list views are a bit more complex, so they will
be adapted in different commits.

Now that the KubeObject has a way to get either a list of the related
resources or an error, it's also very convenient to have a way to get
a message related to that error. Thus, a getErrorMessage is added to
the class so that its implementations can change the message to give
a bit of a better context in the future (if needed).

This new method is very convenient for getting the object list and an
error state. It will be used later for replacing the current use of
useApiList.

We need to be aware when an error occurs when getting a list of cluster
objects, so this patch adds an error callback optional arg to the
mentioned method.

Disabling the Docker cache greatly increases the time it takes to
rebuild an image during development.

The SimpleTable already deals with a "waiting for data" state (when the
passed data is null), but we will need it to be able to display an error
message as well. Thus, this patch adds an errorMessage property for that
effect.

So far we were relying on having the metrics-server installed, so in
clusters where it wasn't available, the related charts would just keep
looping forever (waiting for results to arrive) which is not a great UX.

This patch is a set of changes that will replace the metrics-server
related charts by simpler labels that inform on the capacity only
(rather than the usage) and show a small tooltip icon that informs the
user that they can install the metrics-server for getting more data.

We will need to have a tooltip icon in the mentioned component, so users
may get more information about what the label represents. So this patch
adds that functionality by using an optional property that will add a
TooltipIcon component to the end of the label's title.

This component will be used to display an icon that displays a tooltip
with a message when hovered.

We should include any plugins we want to ship in the image. So, in order
to do that, this patch copies anything in a local plugins folder into
the image and configures the Kubernetes deployment to use this plugins
dir. If the folder doesn't exist, then no plugins will be loaded by the
backend anyway.

In the future we should validate whether the plugins dir has contents
that resemble the plugins (plugins/*/main.js).

Building golangci-lint locally modifies go.mod and go.sum. Committing
these changes prevents a constant diff when running the linter.

This directory contains binaries used for development and should not
be committed.

We should have similar lint rules across our projects, so this patch
adds the same linters as the Lokomotive channel does, and changes the
code so it passes the lint.

For a while we've had multi-cluster support and that means that the
clusters all have a name, even if it's a default one. So this patch
removes the logic for using a prefix that has no cluster name.

Use context auth

Now we support client certificates, which means that the user may not
have to set up a token to authenticate against a cluster. In order to
understand whether the user needs a token, this patch does a quick
auth test in the login dialog, and if it fails then a token will be
required, otherwise the user is redirected to correct view.

This will make it easier to fetch a cluster's info by name which will
be needed when we make the auth token dependent on whether the cluster
needs it or not.

The backend was loading the clusters' info instead of the contexts which
in most cases is fine (as they correspond), but for other cases where a
different user is setup for the same cluster this was not supported.

So, when run with a KubeConfig, this patch loads the information from
the contexts defined there instead of the clusters. This means that we
now take the name from contexts and thus allow the same cluster to be
used twice if defined for different contexts.

The main motivation for this change however, was so that we could use
any client certificates associated with the contexts for accessing the
clusters without needing a token; this is the case for minikube for
example, for which we were requiring the user to set up a token before.
(Changes in the frontend are needed however, so the token requirement
isn't always prompted)

For the in-cluster config, a simple context is created just with the
in-cluster info.

Makefile: Specify Docker build path and ignore node_modules

Sending node_modules to the Docker daemon greatly increases the size
of the Docker build context, which makes `make image` take more time.
We don't need to copy node_modules when building an image because we
run `npm install` in the Dockerfile anyway.

The `docker build` command requires specifying a directory in
addition to the Dockerfile.

Add gitignore

Fix backlinks