test: parallelize suite with ctest

[vulcandth]

Summary

• register the RGBDS test suite as individual CTest cases
• run per-test bash drivers in isolated temp directories so tests can execute safely in parallel
• add labels, smoke-test fixtures, and downstream external tests to support filtering and ordering
• document the new CTest workflow in CONTRIBUTING while keeping the legacy runner available

Metrics

Benchmarks on this machine:

Suite	Old runner	New CTest runner	Speedup
Internal only	1m36s	1m14s	1.30x
External only	2m30s	2m04s	1.21x
Full suite	5m05s	2m37s	1.95x

Additional validation:

• ctest -j$(nproc) -L internal: 728/728 passed
• ctest -j$(nproc) -L external: 11/11 passed
• ctest -j$(nproc): 735/735 passed

Notes

• external tests now run against the fork/root correctly without overriding downstream RGBDS ?= variables
• asm/section-unsigned-overflow gets a longer timeout because it is substantially slower than the rest of the suite
• transient out*.png files are excluded from CTest discovery

[ISSOtm]

Thank you for the idea! There are plenty of benefits to this, as I've outlined in #1927, but this specific implementation duplicates all of our test suite's logic into a separate set of runners without replacing the old ones. Which I'm rejecting on maintainability grounds :P