chore(compiler): Properly handle disable_stamps state

[ospencer]

I noticed that we weren't properly resetting the disable_stamps state in a handful of tests, causing a good chunk of the test suite to not have stamps in their snapshots. For some test suites, this would cause them to fail if run individually.

There's somewhat of a discussion to be had for disabling stamps in a bunch of the snapshots, but I find the stamps really helpful for debugging and think we should keep them. (They really only change when we add new global types.)

[spotandjake]

Closes: #2235