Improve performance of ID generation

[Evangelink]

We have noticed that the ID generation is having an heavy impact on performances of MSTest discovery/run, let's investigate what we can improve and if there are some quick wins let's do them directly.

AB#2050310

[nohwnd]

There are 2 instanced of where IDs are generated for each test, which in total take 20% of execution, plus other interactions with ObjectModel (such as adding traits to traits collection).

I am not sure where the bottleneck of the id generation is, but it does ask reflection for information that we already have (e.g. information about assembly), and that we could pass in. It also does not seem to handle strings in an efficient way.

All the hotspots here are easy to measure by running a test project under PerfView and specifying --list-tests. See the changes here to add stopwatch around the discovery to not take writing to console into account.

https://github.com/nohwnd/testfx/tree/discovery-optimize-and-break

[testplatform-bot]

✅ Successfully linked to Azure Boards work item(s):

• User Story 2050310