feat: Add an assist to extract an expression into a constant

[Giga-Bowser]

closes #7692

Technically, this will permit some invalid const definitions since, as far as I know, we don't have the ability to properly ensure that an expression can be const evaluated, but I believe that would require a larger overhaul of RA's integration of MIR stuff. Either way, this should be fine for now and does a good job rejecting most of the obviously incorrect definitions.

I could also group these assists if that's desirable, just let me know.

[Veykril]

Technically, this will permit some invalid const definitions since, as far as I know, we don't have the ability to properly ensure that an expression can be const evaluated, but I believe that would require a larger overhaul of RA's integration of MIR stuff. Either way, this should be fine for now and does a good job rejecting most of the obviously incorrect definitions.

That's fine, in that case we do want the assist to be more permissive I think. Its okay for assists to generate semantically invalid code (to a certain degree), its not okay for them to generate broken syntax.

[Veykril]

I could also group these assists if that's desirable, just let me know.

This I think we might want to do, grouped together with function extraction, as well as closure extraction, once we have that, as i believe those can trigger on the same selection snippets at times.

[bors]

☔ The latest upstream changes (presumably #18657) made this pull request unmergeable. Please resolve the merge conflicts.

[Giga-Bowser]

This force-push was just fixing merge conflicts, I will begin working on those suggestions now!