Make RISC-V testing reliable

This is a bit of a background task at the moment but the RISC-V testing (currently limited to sanity.openjdk) typically fails due to hitting the 10 hours timeout for the test. I have been experimenting with runnign `PARALLEL=Dynamic` which seems to help lead it to finish. Having said that, the most recent set of runs look like they may have finished underneath the 10 hour limit.

Relevant jobs:
- https://ci.adoptopenjdk.net/job/Test_openjdk19_hs_sanity.openjdk_riscv64_linux/
- https://ci.adoptopenjdk.net/view/Test_openjdk/job/Test_openjdk11_j9_sanity.openjdk_riscv64_linux/ (Competes, but 24/27 targets are skipped)
- https://ci.adoptopenjdk.net/view/Test_openjdk/job/Test_openjdk11_bisheng_sanity.openjdk_riscv64_linux/ (Hits 10 hour limit)

Start by comparing the listed sanity.openjdk RISC-V jobs, especially the OpenJDK 11 J9 and Bisheng runs, with recent completed runs. Investigate the test parallelism and timeout behavior, including PARALLEL=Dynamic; done means the relevant RISC-V sanity tests complete reliably within the 10-hour limit without unexplained skipped targets.

Written by the indexing model from the issue text.