Fix self-hosted test failures

[kanaka]

It's been a while since I've done a comprehensive self-host test run. I've add a input variable to workflow so that self-hosted mode can be added for manually triggered workflows. The CI run is at: https://github.com/kanaka/mal/actions/runs/10309343464/job/28538806985

Some of these implementations might just need a NO_SELF_HOST setting (such as xslt), but a lot of the failures look like bugs rather than timeouts or memory exhaustion.

Once all these have been fixed or NO_SELF_HOST is added, we should make self-hosted test step enabled by default (it only adds a few minutes overall to the full CI workflow).

UPDATE: self-host mode has been made the default for workflows and the remaining broken implementations already had or have been updated with NO_SELF_HOST (erlang, jq, nasm, xslt, vbs).

Here is the the full list of self-hosted failures with very brief statement of the issue (see comments for more details):

• [x] awk
• [x] c.2
• [x] cpp
• [x] common-lisp
• [x] dart
• [ ] erlang - memory leak (e.g. (sumdown 40000))
• [x] java
• [ ] jq - empty lists inside functions/closures are failing
• [x] kotlin
• [ ] nasm - Needs more heap. With increased heap second quasiquote call fails
• [x] ps
• [x] powershell
• [x] python.2
• [x] sml
• [x] vala
• [x] vbs (vbscript) - startup/invocation error
• [ ] xslt - memory/recursion limit
• [x] swift5