Drop the click-completion dependency in favor of the native implementation of click==8.0

[sphuber]

As of click v8.0, tab-completion is now supported natively, as described here in the docs. So far we have been using the click-completion extension which may now be no longer necessary.

There has been an open issue for a while asking whether click-completion can be considered deprecated or whether it still provides more functionality that isn't yet in click. From a first glance, it seems everything we need is there. The only downside is that the completion activation command is now shell dependent and Power shell is not natively supported, but I doubt that this should be a deal breaker.

[chrisjsewell]

Thanks @sphuber, we should look at https://github.com/pallets/click/blob/main/CHANGES.rst#version-800, and have a think if there is anything that either (a) may cause an issue or (b) could additionally improve/replace anything in our current implementation

[chrisjsewell]

When a long option name isn't valid, use difflib to make better suggestions for possible corrections. :issue:1446

could maybe replace our internal implementation

[chrisjsewell]

Use Context.with_resource() to manage resources that would normally be used in a with statement, allowing them to be used across subcommands and callbacks, then cleaned up when the context ends. :pr:1191

this feels like it could have some use, e.g. in #5085 i was thinking it would be “strictly” of the logging config was reset after the cli call (although usually this does not matter, since you only call the CLI for a single “action”)

[sphuber]

Did you close this by accident?

[chrisjsewell]

Oops yeh; the danger of replying on your phone lol