Allow the user to view multiple Kubernetes namespaces at once

Tests relied on url param defaults, which no longer work

Instead of value being "" or omitted meaning "use group.Default", we now allow the empty value
to go through. This was done to allow multi-value options to be able to encode 'no options selected'.
This is important as the alternative is very confusing, as 'nothing selected' would generally mean
'match everything', not 'select default' (which in the k8s namespace example, means the "default" namespace).

Since the UI is the only user of this API, and it already sends the default value always, this does not affect
any existing usage.

Since the UI still wants to show a 'match all' button to prevent confusion, but it's not a normal option
(if it were a normal option, it could be combined with others, which isn't meaningful), we add a new key
NoneLabel that indicates the name that should be displayed on this extra button.

Finally, we implement these changes for the k8s namespace button, ie.
- Get rid of All Namespaces option
- Add "All Namespaces" as the NoneLabel
- Default to "" when "default" namespace not present, which is equivalent to the old "all" option.

This is the first usage of the new 'union' selectType.
Note that we're still sending 'all' for now. There's currently no easy
way to specify this meaning, and arguably it should be done entirely clientside.
But for now it just means some UI weirdness where 'all' is one of the options and having it on
means anything else you select doesn't matter.

This field changes the option group behaviour depending on its value.
Currently only supports two values:
"one" (default): Old behaviour, one option can be selected
"union": Any number of options can be selected, and the filters are OR-ed togther

It is written in such a way as to easily enable a future "intersection" option,
as per union but AND-ing the filters. But this is not done here. YAGNI.

Add version to Scope UI module S3 path

Using `doRequest` when fetching nodes of multiple topologies

Fetch resource view topologies on first topologies load

* Added resource view selector button

* Showing resource boxes in the resource view

* Crude CPU resource view prototype

* Improved the viewMode state logic

* Extracted zooming into a separate wrapper component

* Split the layout selectors between graph-view and resource-view

* Proper zooming logic for the resource view

* Moved all node networks utils to selectors

* Improved the zoom caching logic

* Further refactoring of selectors

* Added sticky labels to the resource boxes

* Added panning translation limits in the resource view

* Renamed GridModeSelector -> ViewModeSelector

* Polished the topology resource view selection logic

* Search bar hidden in the resource view

* Added per-layer topology names to the resource view

* Made metric selectors work for the resource view

* Adjusted the viewport selectors

* Renamed viewport selector to canvas (+ maximal zoom fix)

* Showing more useful metric info in the resource box labels

* Fetching only necessary nodes for the resource view

* Refactored the resource view layer component

* Addressed first batch UI comments (from the Scope meeting)

* Switch to deep zooming transform in the resource view to avoid SVG precision errors

* Renamed and moved resource view components

* Polished all the resource view components

* Changing the available metrics selection

* Improved and polished the state transition logic for the resource view

* Separated zoom limits from the zoom active state

* Renaming and bunch of comments

* Addressed all the UI comments (@davkal + @fons)

* Made graph view selectors independent from resource view selectors

Exclude pause containers when rendering k8s topologies

When k8s present, allow filtering of containers by namespace (take 2)

Map decoding optimisations

'codecgen' embeds a random integer in each identifier; this means code
coverage across different CircleCI lanes may not match.
Here we force the integer to 23 on every CircleCI build so they always match.

This avoids a runtime type lookup, so goes a little faster.
Also having less recursion makes it easier to interpret profiles.

integration test: disable flaky test 311

Temporarily disable the test. See https://github.com/weaveworks/scope/issues/2379

Add job to trigger service-ui build

Clear client polling and nodes state on dismount 

Fix help panel fields max height

Fluent Billing Emitter

Hide 'Save canvas as SVG' button in Table View