* adding .gitignore

* re-ordering items

* fix typos

[ch58083] add close
[ch58076] make connection timeout configuration
[ch58072] add offline mode
[ch58435] make event sending configurable

Linting and some refactoring of streaming code.

A user submitted contribution was merged to support Aeson 2.0. While
all of our tests were passing, this was because we lacked a test
environment that actually used 2.0

This commit addresses the remaining compatibility changes and
introduces a later test environment to ensure we are actually building
with Aeson 2.0 support.

A customer recently provided two pull requests -- one to add the
generated cabal file to our repository and the second to update our
hlint configuration file.

While great, these change aren't sufficient because

- We have no way to enforce the generated cabal file is up to date
- We aren't running hlint during the CI process

This commit introduces CI behaviors to resolve both of these issues.

In a previous commit I updated the releaser config. However, I failed to
adjust two additional bits of configuration.

- The repository now includes a cabal file which contains version
  information. We will now update that.
- docs need to be copied into the appropriate releaser directory for the
  GH pages branch to be updated correctly.

The OSX builds have been taking a very long time, despite doing less
work than the Linux equivalents. A little digging has uncovered
incorrect CI caching for the OSX builds.

This small change has taken the build from approximately 45m to 9m per
OSX run.

A customer contributed this fix. I made some minor tweaks, but testing
seems to indicate everything works as expected.

Co-authored-by: Veronika Romashkina <vrom911@gmail.com>

When we bumped the version numbers for the v3 release, we didn't bump
the corresponding versions in the redis store.

We didn't catch this because we don't try to build the redis package in
CI, and we certainly don't try to build it against every supported Stackage
resolver.

For now, I've taken the simplest solution to get some tests confirming
the behavior. Once I merge this forward in the v4 branch, I will look at
simplifying a lot of the redundancy that exists in our unit tests.