Disable assertion that we are now hitting on enum initializers and file https://github.com/dotnet/roslyn/issues/34919 to follow up.

The private overload was just used by the public one. Makes it easier
to navigate if there's just less code.

I'm guessing these tests originated originally as the Quick Info tests,
and then at some point we broke Quick Info into two pieces: the part
that finds the symbol, and then the conversion of that symbol to the
display form for it. These tests are (nominally) the tests for just the
display service, but it was also testing some Quick Info-specific
behavior making sure we find the symbol in certain edge cases. But it
wasn't doing this by calling into Quick Info: it was doing with a
reimplementation of the first part of Quick Info (namely, finding the
symbol). In short: it wasn't actually testing product behavior, it was
just testing the test's ability to implement its random copy of the
product behavior.

As a result, I'm deleting all the tests out of
SymbolDescriptionServiceTests.vb that were really just testing our
ability to find the symbol; I've verified that there were tests for
the actual product Quick Info for all the scenarios and added the ones
that were missing. This is also good, because it means we have actual
test coverage for product behavior which we didn't before.

This was directly calling internal helpers that I'm potentially going to
have to refactor, and this didn't need any special behavior over our
public helper for this.

Fixes one bug around dropping suppression for a tuple conversion.

Closes #32553

Related issues:

- https://github.com/dotnet/roslyn/issues/34211
- https://github.com/dotnet/cli/issues/10989

.Net - .NET