editor: Fix edit predictions polluting completions menu

[Dnreikronos]

Closes #49565

Before you mark this PR as ready for review, make sure that you have:

• Added a solid test coverage and/or screenshots from doing manual testing
• Done a self-review taking into account security and performance aspects
• Aligned any UI changes with the UI
  checklist

Screenshots
Before:

After:

Release Notes:

• Fixed edit predictions polluting the completions menu with unrelated snippets (e.g. Unicode symbols) when
  show_in_completions_menu is disabled (Word completions in Go files are broken despite "words": "disabled" setting #49565)

[Dnreikronos]

1. Hardcoded true in trigger_completion_on_input: The trigger_in_words parameter was ignored in the default
   match arm, always passing true to open_or_update_completions_menu. This caused word characters to
   unconditionally trigger the completions menu, even when edit predictions were hidden from it.
2. Snippet completions bypassed load_provider_completions gate: Extension snippets (e.g. Unicode symbols like
   "Theta", "tau", "turnstile") were loaded regardless of whether the trigger validation passed. This meant
   snippets polluted the completions menu even when the provider determined completions shouldn't load.

Impact

• Users with show_in_completions_menu: false will no longer see the completions menu triggered on every
  keystroke
• Extension snippets now respect the same trigger validation as other provider completions
• Users who want to remove specific snippets (e.g. Unicode symbols) can still uninstall the relevant extension

[Dnreikronos]

Hi @osyvokon!
I've opened this PR to address the issues described in the issue mentioned above.

[probably-neb]

I think the first change you made is good, seems like a clear bug.

The second change however seems over-constrictive as shown by the test failures on that commit. It prevents any snippets from showing up ever. I pushed up a couple commits to make it so we are a bit smarter about when we show snippets based on matching the trigger against the snippet. Would you test these changes out and let me know if you think it fixes the issue for you?

[Dnreikronos]

I think the first change you made is good, seems like a clear bug.

The second change however seems over-constrictive as shown by the test failures on that commit. It prevents any snippets from showing up ever. I pushed up a couple commits to make it so we are a bit smarter about when we show snippets based on matching the trigger against the snippet. Would you test these changes out and let me know if you think it fixes the issue for you?

Ofc, I will look to that rn!!!

[Dnreikronos]

I think the first change you made is good, seems like a clear bug.

The second change however seems over-constrictive as shown by the test failures on that commit. It prevents any snippets from showing up ever. I pushed up a couple commits to make it so we are a bit smarter about when we show snippets based on matching the trigger against the snippet. Would you test these changes out and let me know if you think it fixes the issue for you?

Seems like that everything is working right: