validator: handle transient errors

[davidpanderson]

A 'transient error' is one that will go away in a while, e.g. an fopen failure because of a broken NSF mount. In general, the BOINC back-end code (validation, assimilation) handles transient errors sensibly:
if there's a bad NSF mount, it retries validation for a few hours rather than marking thousands of jobs as failed.

Extend this to script-based validation.
If a script (either init_result or compare_results) exits with 3, treat that as a transient error.
Treat other nonzero exits (or lack of an exit code) as a permanent error.

More generally (for all validators) add a return value VAL_RESULT_TRANSIENT_ERROR for init_result() and compute_results(). This means any transient error.
Previously we checked only for ERR_OPENDIR.
And for compare_results() we treated all nonzero returns as permanent.

Fixes #5799

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 68 lines in your changes missing coverage. Please review.

Project coverage is 10.49%. Comparing base (b11cd70) to head (d94d625).
Report is 4 commits behind head on master.

Files with missing lines	Patch %	Lines
sched/validate_util2.cpp	0.00%	49 Missing ⚠️
sched/script_validator.cpp	0.00%	19 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #5803      +/-   ##
============================================
- Coverage     10.50%   10.49%   -0.02%     
  Complexity     1068     1068              
============================================
  Files           280      280              
  Lines         35972    36019      +47     
  Branches       8448     8444       -4     
============================================
  Hits           3780     3780              
- Misses        31798    31845      +47     
  Partials        394      394              

Files with missing lines	Coverage Δ
sched/db_purge.cpp	0.00% <ø> (ø)
sched/script_validator.cpp	0.00% <0.00%> (ø)
sched/validate_util2.cpp	0.00% <0.00%> (ø)