refactor: Allocate IDs for tests

[nayeemrmn]

Blocked on #14757.

Allows for better state-keeping of tests in Rust:

• Prints ... cancelled for tests that were registered and cancelled due to uncaught errors:

  running 3 tests from ./temp.ts
  testWithUncaughtRejection ...
  Uncaught error from ./temp.ts FAILED
  testWithUncaughtRejection ... cancelled (0ms)
  anotherTest ... cancelled (0ms)
  anotherOne ... cancelled (0ms)
  

• Always know whether or not to reprint someTest ... when a result arrives (addresses Deno.test with output: Show ok on the same line as the test name #14315 (comment)).
• Allows lots of future improvements to output e.g. more self-contained result lines for concurrent runs:

  (./temp.ts) t1 ... ok
  (./temp.ts) t2 ... t2-step1 ... FAILED
  (./temp.ts) t2 ... t2-step2 ... ok
  (./temp.ts) t2 ... FAILED
  

Adds op_register_test() to register tests, allocate IDs to them and potentially other info:

• Move filtering to rust, { id, filteredOut } is returned by this op.
• Tests are known about in Rust earlier, allowing for the ... cancelled output above for tests that haven't started yet.

Cleanups in LSP test runner and more functional approach in 40_testing.js. Bring bench runner error formatting up to date, it was still using formatError() in JS.

[dsherret]

What happens in the vscode's testing sidebar if you run a test with steps multiple times?

[dsherret]

Seems to work.

[dsherret]

LGTM. Probably @kitsonk should review the LSP changes before merge though.

[bartlomieju]

LGTM, nice work Nayeem. I verified that LSP changes work properly in vscode_deno.

[kitsonk]

I'm concerned some of the changes to the lsp aren't "cleanup" but arbitrary changes. Specifically for the messages that go to and from the client that are changed should have their types changed in https://github.com/denoland/vscode_deno/blob/main/client/src/lsp_extensions.ts to ensure there isn't a knock on effect. Especially when we remove Option it becomes a mandatory field on deserializing.

[dsherret]

@nayeemrmn seems like this is flaky now: https://github.com/denoland/deno/runs/7362087847?check_suite_focus=true