High availability edition/database is deprecated by this PR in favour of
causal clustering.
All high availability related public/not public code will be removed
in next major release. All existent tools and usages should be updated
to use causal clustering instead.

By closing allocators properly we clearing allocated memory chunks
and that reduce time we that we need to spend to perform pointer ranges
checks.

We still don't understand some of the performance implications of compiled
expressions. So until we do, turn them off by default.

This was completely removed in 3.4 anyway and this way we don't need to care about a bunch of old dependencies

Marked internal, defaults and descriptions copied over from Akka config.

Also do a slight bit of clean up.

This allows us to defend against misalignment between the in row property
values and any property write which could occur during the query. Special
runtime handling of cached node properties let's us check whether the
property for that node has been changed in this transaction, and in that
case use the tx-state value instead of the in-row value.

that used at least in bolt server.

Encapsulate base edition module fields. Small update on how
security module is build
(removing static initialization is our of scope of current PR).

Use creation context instead of edition directly for database creation
since it simplifies lifecycle and dependency management of some editions
that do that for their internal components.