PromiseRegistry for caching async work

[markfields]

When cleaning up promise-function-async disables under the drivers dir, there were two cases I left in place, because the functions were about caching/retrieving Promises, and needed to stay synchronous to discourage accidentally awaiting the promises before the caching was resolved properly. After finding a third similar pattern (and 2 bugs) I played around a bit and came up with a way to refactor the tricky parts of that logic out.

See those 3 places updated to use PromiseRegistry in the draft PR #1713.

Once this is merged and the client packages are consuming fluid-common-utils version 0.15, I'll put out another PR to add those 3 usages.

To-do:

• Fix build break (my local build has been pretty messed up...)
• Write a million tests

[markfields]

Well, I'm actually trying a different API surface, modeling after .NET's ObjectCache. I think it is easier to understand, based on the precedent of the .NET API and also some feedback on the draft PR that I've slowly digested.

[arinwt]

Looks good. Depending on how it looked, I might have made the "indefinite" form a class on its own with the expiration forms in an class extending it, but definitely not necessary since this is already fairly simple.

[markfields]

Update: Refactored GC into a helper class and wrapped asyncFn() in an async lambda to ensure it doesn't throw.

[markfields]

50% more LOC in the tests than the implementation itself 😛