* Refactor tuple implementation

* Tweaks

* Address PR feedback

* Pass tuple data in constructors

* Address more feedback

* Tweak comparison

* Restore constructWithTypeParameters flag to avoid overflow

* Assert that tuple parameter matches TupleUnderlyingType somewhat

* Rename to TupleData. Address some IDE feedback

* Remaining TODO2's

* Fix last IDE test!

* Factor array comparison

* Tweak WithTupleData and EqualsIgnoringTupleUnderlyingType

* Expand a test

* Remove duplicate line

* Remove an EE test for badly formed ValueTuple type

* PR feedback from Neal

* Restore an IDE test without ValueTuple

* Rename constants

* Restore formatting

Delete a bunch of code on Roslyn side, which is no longer necessary as we don't own the column anymore.
I have also cleanup up some code in this space for newly added ContainingType and ContainingMember columns.

* Both columns added

* Genericize custom column names

* More changes to genericize

* Cleanup

* Respond to feedback

* WIP

* Change tuple to properties

* Restrict to just containing types

* Cleanup

* cleanup

* Refactor

* Change method for getting Containing Type and Member

* refactor GetAdditionalProperties

* Fix merge conflicts

* respond to feedback

* respond to feedback

* Change to struct

* Add type for AbstractCustomColumnDefinitionWithMultipleValues

* Cleanup

* Change AdditionalProperty to FindUsagesProperty

* respond to feedback

* respond to feedback

* Hide columns in F#

* respond to feedback

* fix merge conflict

* Change from hiding columns to removing columns for non supported langs

* Add additional implementations

* formatting

* removed IEditorClassificationService that is marked obsolete since 2017 and bunch of other code that only existed for the IEditorClassificationService.

according to vso, only F# still has dependency so I talked to them to move to new API (IClassificationService) - https://devdiv.visualstudio.com/DevDiv/_git/DotNet-Source-Build-Tarball?path=%2Fsrc%2Ffsharp%2Fvsintegration%2Fsrc%2FFSharp.Editor%2FClassification%2FColorizationService.fs&version=GBmaster

and a bunch of clean up and move code that has no dependency to the editor to lower layers so that I can consume in razor scenario.

* added shim project for FSharp that has internal visible to FSharp editor

* merged https://github.com/dotnet/roslyn/pull/31941

* addressing PR feedback

* moved some type (such as DocumentSpan, ClassifiedSpan and etc) back to where fsharp has IVT to.

can't move those since partner team is using them and they don't have IVT to workspace layer for some reasons.

also, have put the IEditorClassificationService back and add a proxy in FSharp External Access so that it can still work until they remove all IVT.

- Include non-abstract definition into the implementations result.
- Include implementations from derived interfaces into the result.

Implementation of the custom column support has been kept extensible, so that each reference item can have customized key-values info, where key corresponds to the name of the custom dynamic column and values are set of one or more values the table entry. Columns become dynamically visible if at least one row in the table has a value for it.

Fixes #24877

* fix quickinfo close brace

* PR comments

* rename

* PR comments

* PR comments

* fix new line leading whitespace

* AddRange

* remove the Assert not null

* Extract code for classifying a document span into its own helper class.

* Move file.

* use EditorClassifier.GetClassifiedSpansAsync and IndentationHelper.GetSpansWithAlignedIndentation

When we were navigating to a symbol with Go to Definition, we were
roundtripping the symbol to a symbol key, and then trying to
reconstitute the symbol by finding a project that had the reference
and then grabbing that symbol. There are two problems with this
approach:

1. We might get the symbol for the wrong language, which impacts the
   language we use to generate metadata as source. This fixes
   dotnet/roslyn#16818.

2. We might not even get the right symbol. Assembly identity in this
   case was using display name, which could be the wrong portable
   surface area or wrong version.

The fix is to instead carry along the project ID of the originating
symbol, and resolve against that.