Simplify completion helper code.

Remove turkish specialization.

Fix a null reference in BaseFormattingRule...

Make 'SplitStringLiteral' fully synchronous.

Fix FxCop warnings

... that was introduced with PR #4644

Method `SomeParentHasMissingCloseBrace` expects a fully formed node that
has a compilation unit, which a roslyn document would have, but this
code is also in a formatter public api path, that can possibly `an
arbitrary` syntax node that may not be associated with a syntaxtree or a
compilation unit.

The fix is simple, it adds a missing null check to this method.

Update nonshipping flags

add <ShortcutPath> and NGEN info to VSIX packages

It's contained in the diagnostics window VSIX, which we include in
the full deployment.

This refactors our SignVsixInputs task into two pieces, one that
simply computes the list (so we can validate it) and then the other
piece that actually does the signing. The first bit runs always, even
on non-signed builds.

Here's why we need this:  This set of commits changes SplitString and Indentation to use a synchronous
path for getting a syntax tree.  This is so that those features willblock hte UI thread for as small an
amount of time as possible.

*However*, it's not always the case that they actually get the syntax tree synchronously.  in order for
them to get the tree, they need to call AsyncLazy.GetValue to actually get the value.  However, AsyncLazy.GetValue
will not actually perform synchronously if there is an existing async request already in flight.

In practice that can happen because the background parser ends up kicking off work to try to get the
syntax tree asynchronously.  Because of this, when we call into get the tree synchronously the asynclazy
actually has us wait on the async work that is in flight.

In tests this causes us to assert because we're blocking on async work on a non-UI thread.  We had two options
here:

1. Change AsyncLazy to not wait on the async work here and instead go execute the work synchronously.
2. Change teh tests to be WpfFact

Jason and I discussed hte first option but decided against it.  Changing async lazy has huge downstream ramifications
and we weren't comfortable just changing it's behavior in that manner.

Note: during tests we found 99% of the time, we did go through the sync path.  So, ideally, the impact
of us blocking the UI thread on async work is minimal in practice.

Merge future into master

Merge stabilization into master

Adding back the PopulateSwitch tests.

Delete dead code.

Provide a synchronous path for outlining so it won't block hte UI thread waiting for a threadpool thread.

Fix tests.

Provide a synchronous path for outlining so it won't block hte UI thread waiting for a threadpool thread.

mrege conflict resulting from changing CreateCodeRefactoringProvider return type from
object to CodeRefactoringProvider

* initial port from CSharpEssentials

* responding to feedback.
removing syntax rewriter and incorporating formatting rules into formatter

* responding to cryus' feedback

- general logic cleanup
- moving methods to abstract base class

Merge future into master