runtests: enable torture testing with threaded resolver

[bagder]

Since a7bebd8 made it possible.

Also enable memdebug in test 3207 again.

[testclutch]

Analysis of PR #19786 at 89acdb36:

Test ../../tests/http/test_16_info.py::TestInfo::test_16_03_info_upload[h2] failed, which has NOT been flaky recently, so there could be a real issue in this PR. Note that this test has failed in 2 different CI jobs (the link just goes to one of them).

Generated by Testclutch

[vszakats]

It seems that Windows arm64 tests became much slower
and can no longer fit in 10 minutes with this patch. Before,
it finished in 5-6 minutes.

First seen in this PR, and in almost all runs since:
https://github.com/curl/curl/actions/runs/19823354249

It also applies to Intel builds, but those took 3 minutes and
now 6, which still fits. But also much slower.

Reference from before:
https://github.com/curl/curl/actions/runs/19797266382

master build after:
https://github.com/curl/curl/actions/runs/19824903607

(affects both MSVC and mingw. there is bottleneck somewhere
on Windows apparently.)

[vszakats]

windows arm64 MSVC time summaries:

Before:

  Test suite total running time breakdown per task...
  
  Spent 0946.949 seconds starting and verifying test harness servers.
  Spent 0091.779 seconds reading definitions and doing test preparations.
  Spent 0927.179 seconds actually running test tools.
  Spent 0214.141 seconds awaiting server logs lock removal.
  Spent 0028.763 seconds verifying test results.
  Spent 2208.811 seconds doing all of the above.

After:

  Test suite total running time breakdown per task...
  
  Spent 0849.428 seconds starting and verifying test harness servers.
  Spent 0632.870 seconds reading definitions and doing test preparations.
  Spent 0923.245 seconds actually running test tools.
  Spent 0127.122 seconds awaiting server logs lock removal.
  Spent 0974.010 seconds verifying test results.
  Spent 3506.675 seconds doing all of the above.