[clang-tools-extra] [clangd] Fix crash on completion with out-of-range position (PR #196112)

[Aleksandr Platonov via cfe-commits]

ArcsinX wrote:

> The reason I haven't changed the result of `shouldRunCompletion` is twofold:
> 
> 1. If we return false, we'll just reply with empty result as you noticed, logging `vlog("ignored auto-triggered completion, preceding char did not match");`. I'm not sure this is the intended behaviour of the API - I assumed passing invalid location should be reported to the client as error as it shouldn't really happen. Unless we change the signature of `shouldRunCompletion` we can't really trigger error response from there.
> 2. There's already precedent for letting the completion code handle error reporting in the line before (the `getDraft` call can return nullptr - we still return `true` there knowing the completion code will return error to the client).
> 
> Would we rather prefer silently (from client's perspective) ignoring the invalid position?

Sorry, maybe I wasn't clear enough.
I don't see any difference in behavior if we return `false` in `shouldRunCompletion`. We'll still return
```json
{
  "isIncomplete": false,
  "items": []
}
```
Just like if we returned `true` from `shouldRunCompletion`.
So, from LSP standpoint we don't notify the client of an error, we just return an empty result.

The only difference is in the log output. Therefore, I suggest returning `false` and replacing `vlog` with `elog`. Then, from a diagnostic standpoint, everything will remain the same, but we return empty result earlier.

https://github.com/llvm/llvm-project/pull/196112