Add a regression test for completion list in object literal involving inferred obj with optional members

[Andarist]

This is just a regression test.

I was investigating a case that I've discovered some time ago but I discovered that it got fixed in 4.6. So I've decided to check if my issue was fixed directly or if, perhaps, it was fixed accidentally - and in general, I've wanted to check how it was fixed ;p

I was able to narrow down the fix to this PR: https://github.com/microsoft/TypeScript/pull/46743/files and I've confirmed locally that removing this single line of code brings the issue back. As we can see - the fixed issue by this PR is a different one, so I think that it's worth adding my case as a regression PR.

I've prepared a TS 4.5 playground demonstrating the issue. You can compare it with the TS 4.6 playground where the issue is gone.

I don't really understand the fix - didn't have time to dig into this deeper. Perhaps @a-tarasyuk or @andrewbranch would understand why does it help for this case too.

[typescript-bot]

This PR doesn't have any linked issues. Please open an issue that references this PR. From there we can discuss and prioritise.

[andrewbranch]

Synthesized mapped type member symbols don’t have declarations which is why they’re affected by that fix. The function changed in the fix attempts to filter out completions for f({ blah/**/ where the blah you’re typing is the only place that causes a blah member to exist.

It seems we’ll be unable to determine that when the object literal member is being inferred to a mapped type, so while the fix made things better for hover and pressed, I would expect that it makes things worse for additional keys, giving a completion item that simply matches whatever you’re currently typing.

But checking, that doesn’t seem to be the case:

So I don’t know exactly what’s going on there, but I’m glad it’s all working 😄