Reduce non-highmem browser testing

[dschuff]

This reduces the number of tests run in the "default" browser test modes
(i.e. non-2g 32bit and non-4g 64-bit). The idea is that the set of edge
cases covered by the 2g and 4g modes is basically a superset of those
covered by the default cases. Tests that are disabled in the 2g and 4g
modes are still run in the default modes.

[dschuff]

This does still keep some tests running in the default modes, just to make sure that no tests are skipped entirely.
It's not awesome that the list of tests that we still run in the default mode is duplicated manually from the set of tests marked with no_2gb and no_4gb. Maybe if we want to keep this optimization longer-term we could do something about that, not sure.

[sbc100]

LGTM with comments and maybe TODO to restore some day?

[dschuff]

test_2gb_fail, test_4gb_fail and test_2gb are failing here in test-browser-chrome-wasm64. Were they working before?

[dschuff]

oh nevermind, they were skipped before.

[sbc100]

test_2gb_fail, test_4gb_fail and test_2gb are failing here in test-browser-chrome-wasm64. Were they working before?

The 4gb_fail test looks like it was deliberately skipped (your removed that line).