Add InvalidHandleValue metadata for REGHANDLE

[riverar]

From evtprov.h:

/*
A REGHANDLE represents an event provider session.
A REGHANDLE is opened with EventRegister and closed with EventUnregister.
*/
typedef ULONGLONG REGHANDLE, *PREGHANDLE;

From docs:

If EventRegister fails, it returns a zero-valued RegHandle. ETW APIs such as EventWrite and EventUnregister will safely accept a zero-valued RegHandle value, do nothing, and return immediately. Callers do not need to check for a zero-valued RegHandle before calling EventWrite or EventUnregister.

Suggested fixes:

1. Add [InvalidHandleValue(0)]
2. Add metadata test to fail if IHV metadata is missing from RAIIFree marked types.

Originally posted by @kennykerr in #1891

[riverar]

Additional:

• BSTR (⚠️ This is not a handle type)
• HSTRING (⚠️ Has no invalid handle value)
• PROCESSTRACE_HANDLE
• PDH_HLOG
• PDH_HQUERY
• PRINTER_HANDLE
• LPPROC_THREAD_ATTRIBUTE_LIST (⚠️ This is not a handle type)

[kennykerr]

This is not a handle type

I was curious what you meant by this (I'm still curious) and then I happened to notice that there seems to be a new metadata attribute I wasn't aware of. HSTRING for example has the NativeTypedef attribute while PROCESSTRACE_HANDLE has the MetadataTypedef attribute. Rust only deals with the former, but the latter seems to indicate the same or similar thing. Why are there two attributes?

[riverar]

I was just making a note to indicate which types shouldn't get an InvalidHandleValue assigned to them, given that attribute is meant for handles and not pointers. I recognize HSTRING is a handle; HSTRING is notable though as it has no invalid value.

NativeTypedef indicates the typedef declaration natively appears in the headers, while MetadataTypedef indicates it was artificially created.