don't free alloc-bypass resources when scheduler reloads

[garlick]

This fixes #5797 by omitting alloc-bypass jobs from the "hello" phase of the scheduler handshake.

It looks like set-flags is already used to ensure job->alloc_bypass survives a flux restsart.

I did notice that my flux restart test hangs without FLUX_DISABLE_JOB_CLEANUP set, as the alloc-bypass job gets stuck in CLEANUP, so that needs following up. Will open a separate issue.

[garlick]

I added a fix for #5981

[garlick]

Thanks! I'll set MWP.

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 83.29%. Comparing base (7bd3e35) to head (a7c0ffa). Report is 488 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5980      +/-   ##
==========================================
- Coverage   83.29%   83.29%   -0.01%     
==========================================
  Files         515      515              
  Lines       83405    83399       -6     
==========================================
- Hits        69475    69467       -8     
- Misses      13930    13932       +2     

Files	Coverage Δ
src/modules/job-manager/alloc.c	75.91% <100.00%> (ø)
src/modules/job-manager/event.c	78.57% <100.00%> (+0.04%)	:arrow_up:
src/modules/job-manager/plugins/alloc-bypass.c	67.50% <ø> (-0.32%)	:arrow_down:

... and 12 files with indirect coverage changes