* `buffers` is an unnecessary intermediate variable. Any text buffer present in `buffersToTag`
  but not also present in `buffers` will end up ignored by `ComputeNewTagTree`, and thus not
  influence the result of `ConvertToTagTrees`. Based on this, `buffers` was eliminated to
  simplify the implementation.
* When `spansTagged` has only a single span, it's possible that other tag trees in `oldTagTrees`
  need to propagate forward. The early return failed to account for this, and if hit would drop
  `oldTagTrees` tags for buffers not equal to the buffer from `spansTagged`. In addition, the
  early return failed to account for the case where `spansTagged` only contained a single
  span from a buffer not requested per `buffersToTag`. Following the elimination of `buffers`,
  the value of a special case handling here was notably reduced and the edge case was simply
  eliminated in favor of the remaining code in the method.
* The patter of Select followed by ToLookup was replaced by a single call to ToLookup specifying
  both the keys and the values of the resulting lookup.

Fixes #25255
Fixes #26329

Fixes #25244
Fixes #25731
Fixes #27187

I used `\n` and `\r\n` in string instead `@""` because as far as I understand on Windows/CI code is checked out with `git config core.autocrlf` which means it modifies line endings...

* remove usings when format document

* sort usings when format document

* add code fix

* hack fix for the issue with apply multiple changes.

* refactoring

* 3 changes

1. put open file editorconfig change tracking back
2. handle open file leaking when solution close with files opened case
3. don't share locks between editorconfig events and roslyn events

* skip the 'Remove unused variable' when format document

* changed code fix part to use "fix all" and only fixes P1 list rather than fix everything in the document. also not apply changes to buffer until everything is done rather than doing it per fix.

* made undo experience better

* add expression body as one of fix to offer

* add code fix options in format document settings

* add hidden box IsCodeCleanupConfiguredCheckBox to track if the user ever configured the code cleanup

* clean up

* fix some tests

* disable a test temporarily

* PR comments

* refactoring

* Revert "3 changes"

This reverts commit 7e0fd53f02a0e20d063440b3cce5569e51e71afc.

* PR comments

* PR comments

* renaming

* bug fixing

* PR comments

* refactoring

* refactoring

* refactoring

* refactoring

* Make ICodeCleanupService extend ILanguageService

* refactoring

* refactoring

* fix deadlock, call Format inside CodeCleanupService

* rename async methods

* PR comments

* move files under features

* PR comments

* PR comments

* Never Show this again will never show the code cleanup info bar for this user.

* rename

* use ImmutableArray

MonoDevelop(VSfM) recently switched from using `Microsoft.CodeAnalysis.Classification.Classifier` to using `IAccurateClassifier` from VS API.
But with this switch some classifications that were added in #24931 stopped working. Reason for this is that in that commit new classifications were filtered out by converting back to classification that was used before `ClassificationTypeNames.Identifier` so VS Windows would continue working fine. But later in #25723 this ClassificationTypes were added so I assume with removing this filtering VS Windows will work just fine since ClassificationTypes that were added #25723 have BaseType `ClassificationTypeNames.Identifier`.

We had lots of places doing GetTextAsync() and then blocking, which is
an anti-pattern if we can avoid it. Awhile back we plumbed the ability
to get syntax trees synchronusly in certain places, and so we had
the plumbing to do this. We just never did it.

I suspect "most cases" this was fine because the GetTextAsync was
operating on open files which isn't an issue. But there were a few
places that looked suspicious and it's easier to just do a batch
refactoring.

* guard us from completely failing when classification type map is messed up.

also add NFW to save dump on such situation

* PR feedbacks and made us to guard from classification issue.

* addressed PR feedback

* PR feedback. add back NFW

Cancellation allows item removal from the pending work queue from non-UI threads,
so the UI thread handling needs to account for the possibility of it being empty
after WaitForItemsAsync completes.

This reverts commit f56d67a6.
We are seeing deadlocks where our code tries to fetch the options
for a file, and that code runs on the background thread. The background
thread is trying to hop back to the foreground thread to apply file
watchers for .editorconfig files.

Leading/trailing decorative text can interfere with the pattern
matcher's ability to find/bold matching spans and add redundant items to
the MRU list.

* Update Roslyn snapshot when editorconfig is updated. for now, it only updates when editorconfig is saved.

* added more comments

* addressed PR feedbacks

Provide an indirection so that individual langauge service features don't have to know about each specific embedded language we create.