[emscripten] Assorted fixes and test enhancements

Tests:
- Parameterize the tests so we don't have to duplicate tests between
  core and compat. Most of the tests test exactly the same thing in both
  cases but the total runtime is still < 1s so it's fine.
- Make the compat tests mostly work on browsers without compat (except
  for one targeted test). Tested on Firefox and Safari.
- Test that all limits are initialized.
  - This would have caught the bug fixed in
    https://dawn-review.googlesource.com/c/dawn/+/294495
    (independently of the externs fix that also would have caught it).
- Test that all limits are non-zero except for maxImmediateSize.
- Make DeviceLevelTests.ValidationError time out, so it will fail
  instead of hang on Safari.

Implementation:
- Make compat limits fall back to their core equivalents for browsers
  that don't implement them yet.
- Skip passing through the new limits (that we added to Core for Compat)
  to browsers that don't implement them yet.
- Fix missing maxBindingsPerBindGroup in emwgpuAdapterRequestDevice.
  (Didn't catch this before because Chromium only exposes 1000, but
  Firefox and Safari expose 65535 so the test failed.)
- Move maxImmediateSize up next to the other core limits (rather
  than buried below the compat limits), and format them so it's easy to
  tell where they differ.
- Remove theoretically unnecessary check for maxImmediateSize==undefined
  that was added in https://dawn-review.googlesource.com/c/dawn/+/242374
  but unfortunately the logs from that original failure no longer exist,
  so I can't verify if that issue would still happen or not.

Bug: none
Change-Id: Ieab362a76f70804d4f3581f7dde678ad38ea35ea
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/294516
Reviewed-by: Loko Kung <lokokung@google.com>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>