Fixes #37213

1. Usages as an argument to `nameof`: We have multiple reports of these being false positives in presence of reflection based usages and in variour test frameworks. Fixes #32488.
2. Presence of 'dynamic' in C# - Conservatively bail out in presence of any dynamic operations. Fixes #32522.

Found while dogfooding latest bits

Detect references to method/property in bound method group/property group using semantic model APIs. Also provide a different message for methods which have name only references.

Fixes #30895

This analyzer does not function correctly in presence of VB handles clause due to IOperation bugs #30978 and #30979. This change adds a temporary workaround to the analyzer to handle this case specially.

TODO: File a bug to track reverting the temporary workaround once #30978 and #30979 are fixed.

Add an explanatory comment and revert the semantics of the new TestParameters flag so the default value (false) retains the current semantics.

Address PR feedback (use static ctor and remove regions) and also fix an assert seen during dogfooding.

Primarily addresses https://github.com/dotnet/roslyn/issues/29519#issuecomment-420067012

```
1. Enable the rules be default
2. Remove all the workaround/hacks for disabling the rules for product and enabling for tests.
3. Address offline feedback from @sharwell about avoiding deriving the tests from `TestBase`.
4. Separate out the pending non-design related work items from first comment into separate issues.
```

For 4., I have repurposed https://github.com/dotnet/roslyn/issues/29519 to track the pending issues with checkboxes in the first issue comment.

Address some more feedback: Report unread member instead of unused member for fields with non-constant initializers.

1. Break the analyzer into two separate diagnostics for flagging unused and unread members. Latter rule has no code fix.
2. Remove all UI option elements for the added diagnostics and turn the rules off by default for now.
3. Bail out for invalid operations and syntax errors.
4. Handle doc comments and nameof - report unread member diagostics instead of unused member diagnostics if referenced in non-read/write context.

Analyzer flags two cases: members with no read/writes and members with only writes.

1. Members with no read or writes: `Type '{0}' has an unused private member '{1}' which can be removed.`
2. Only writes: `Type '{0}' has a private member '{1}' which can be removed as the value assigned to it is never used.`

Code fix removes the unused member declaration.
Fixes #24225

Open questions:
1. Current analyzer design uses a single code style option for all members and both the above kinds of unused members. We can potentially have multiple options, but this should probably be done based on feedback.
2. Should the analyzer use different diagnostic IDs for the above two kinds of unused members? This will mean that the FixAll experience will need multiple iterations for removing unused members.
3. Should we update the code fix (or have an additional code fix) that also updates the write references for the case (2) above? Or is it better to leave the references to break the code so the user can analyze if the value being assigned can also be removed. The PR current chooses the latter approach as it is more conservative and unlikely to cause silent breaks.