[ActivityInfo] Avoid writing extra byte by updating length

[mdh1418]

Fixes #109078

It was discovered that in some scenarios, the ActivityTracker.AddIdToGuid method's optimization logic fails to update the number of encoded bytes counter for (0xFF, 0xFFF] IDs. As a result, an extra 0x00 byte is encoded in the Guid, which is interpreted as the end of the list.

This PR updates the len variable accordingly, so only one byte is written after bundling the (0xFF, 0xFFF] ID's 4 high order bits into the multicode indicating byte.

Before
1/4/2000/1 -> 14 c7 d0 00 10

After
1/4/2000/1 -> 14 c7 d0 10

[mdh1418]

It seems like the GUID optimization logic (squeeze as much info into as few bits as possible by coupling an odd-numbered nibble Id's highest order nibble with the multicode indicator nibble) only occurs for IDs in [0x0, 0xF] u (0xFF, 0xFFF], but not for (0xFFFF, 0xFFFFF].

Is it worth extending the optimization to (0xFFFF, 0xFFFFF] Ids or do Ids not reach those values?

[noahfalk]

Is it worth extending the optimization to (0xFFFF, 0xFFFFF] Ids or do Ids not reach those values?

I wouldn't worry about extending it further. I don't recall ever seeing a complaint related to it and if we did extend it that would break compat with PerfView's decoding of the current scheme.