Don't crash h_abstraction() on empty product_dicts

[alongd]

The crash

h_abstraction() in arc/job/adapters/ts/heuristics.py indexes
reaction.product_dicts[0] before the loop that iterates over the same
list:

xyz_guesses = list()
dihedral_increment = dihedral_increment or DIHEDRAL_INCREMENT
reactants_reversed, products_reversed = are_h_abs_wells_reversed(rxn=reaction, product_dict=reaction.product_dicts[0])
for product_dict in reaction.product_dicts:
    ...

When reaction.product_dicts is empty (no matching H_Abstraction family
products were identified for the reaction), the for loop below would
simply not execute and leave xyz_guesses empty — a correct "no guesses"
outcome. But the unguarded [0] on the preceding line raises
IndexError: list index out of range first, turning "no guesses" into a
fatal crash.

Real-run evidence

Observed in a multi-hour T3 PDep→QM ARC run. The exception propagated all
the way up through Scheduler.__init__ and killed the whole run:

  File "/home/alon/Code/ARC/arc/scheduler.py", line 557, in __init__
    self.schedule_jobs()
  File "/home/alon/Code/ARC/arc/scheduler.py", line 633, in schedule_jobs
    self.spawn_ts_jobs()  # If all reactants/products are already known (Arkane yml or restart), spawn TS searches.
  File "/home/alon/Code/ARC/arc/scheduler.py", line 1795, in spawn_ts_jobs
    self.run_job(job_type='tsg', ...)
  File "/home/alon/Code/ARC/arc/scheduler.py", line 1072, in run_job
    job.execute()
  File "/home/alon/Code/ARC/arc/job/adapter.py", line 229, in execute
    self.execute_incore()
  File "/home/alon/Code/ARC/arc/job/adapters/ts/heuristics.py", line 266, in execute_incore
    xyzs = h_abstraction(reaction=rxn, dihedral_increment=self.dihedral_increment)
  File "/home/alon/Code/ARC/arc/job/adapters/ts/heuristics.py", line 879, in h_abstraction
    reactants_reversed, products_reversed = are_h_abs_wells_reversed(rxn=reaction, product_dict=reaction.product_dicts[0])
IndexError: list index out of range

Root cause of empty product_dicts

ARCReaction.product_dicts lazily calls get_product_dicts() →
get_reaction_family_products(), which returns an empty list whenever no
RMG reaction-family template matches the reaction (e.g. atom-mapping
ambiguity, an unusual/edge-case reaction, or a family that legitimately
doesn't apply). This is an expected, already-handled emptiness elsewhere in
the file: get_products_and_check_families() (around line 981) has its own
if not product_dicts: branch for the case where a specifically-requested
family produced no products. There's no evidence of a deeper upstream defect
here — a TS-guess heuristic simply has to tolerate "no guesses could be
constructed" as a valid outcome rather than treating it as fatal, since ARC
has other TS search methods to fall back on.

Fix

Guard the [0] access: if reaction.product_dicts is empty, log a warning
naming the reaction and return an empty guess list immediately, before ever
touching [0]. This matches the existing logging/docstring style in the
file (logger.warning, numpydoc-ish Returns section).

Sibling-pattern search

Searched the rest of heuristics.py for the same "unguarded [0]-index
before a loop over the same list" pattern in the other family heuristic
functions (hydrolysis, get_products_and_check_families,
get_reaction_family_products, etc.). None found — h_abstraction() was
the only function with this defect.

Tests

Added test_h_abstraction_with_empty_product_dicts to
arc/job/adapters/ts/heuristics_test.py, which builds a real H_Abstraction
ARCReaction, forces product_dicts = [] via the existing setter, calls
h_abstraction() directly, and asserts it returns [] instead of raising.

Baseline (before fix): 43 passed (heuristics_test.py)
After fix: 44 passed (heuristics_test.py)

Test plan

• python -m pytest arc/job/adapters/ts/heuristics_test.py -q — 44 passed (was 43 before this change; new test added)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 64.23%. Comparing base (06c6ce6) to head (df98d7b).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #993      +/-   ##
==========================================
+ Coverage   64.20%   64.23%   +0.02%     
==========================================
  Files         119      119              
  Lines       39578    39580       +2     
  Branches    10266    10266              
==========================================
+ Hits        25412    25424      +12     
+ Misses      11196    11186      -10     
  Partials     2970     2970              

Flag	Coverage Δ
functionaltests	64.23% <ø> (+0.02%)	⬆️
unittests	64.23% <ø> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.