Auto-detected bugs - Classifications - hard ones

Split from #1446 these are not obvious and need thought

--------------------------------------------------------------------------------

BUG 5 — is_pending always returns True (type mismatch comparison)
FILE: classification/views/discordance_report_views.py
LINE: 102
IMPACT: HIGH — pending-change indicator is always True, breaking the
discordance report UI's pending change display

```
@dataclass(frozen=True)
class _LabClinSig:
lab_clin_sig_key: _LabClinSigKey # a frozen dataclass
count: int
pending_clin_sig: Optional[str] = None # a string or None

@property
def is_pending(self) -> bool:
return self.lab_clin_sig_key != self.pending_clin_sig # BUG
```

A _LabClinSigKey object will never equal an Optional[str], so this property
always returns True regardless of whether there is actually a pending change.

Likely intended comparison (checking if a pending change to a *different*
classification exists):
return self.pending_clin_sig is not None and self.pending_clin_sig != self.clin_sig

Two plausible fixes:
- return self.pending_clin_sig is not None — "there is any pending change at all"
- return self.pending_clin_sig is not None and self.pending_clin_sig != self.clin_sig — "pending change targets a different classification"

Note: is_pending is currently never called anywhere (not in templates, not in Python). The template accesses pending_clin_sig directly. So this is dead code right now
— needs a decision on whether to fix it for future use or delete it.

--------------------------------------------------------------------------------

BUG 6 — "leave it as it was" comment but code overwrites the value
FILE: classification/views/discordance_report_views.py
LINES: 267–269
IMPACT: MEDIUM — when multiple pending clin-sig directions exist, the
suggested_pending_cs ends up as key.clin_sig (current value) instead
of remaining as the first-detected pending target

```
for pending_cs, pending_count in pendings.items():
total_count += pending_count
if suggested_pending_cs:
# can't handle pending in multiple directions, just leave it as it was
suggested_pending_cs = key.clin_sig # <-- BUG: overwrites, not "leaves"
else:
suggested_pending_cs = pending_cs
```

The comment says "leave it as it was" but the code replaces suggested_pending_cs
with key.clin_sig (the *current* clin sig, not the already-set pending target).
Should be `pass` (do nothing) to actually leave the variable unchanged.

Three plausible fixes when multiple pending directions are found:
- pass — keeps the first pending direction found (matches the comment literally)
- suggested_pending_cs = None — clears it to signal "ambiguous, show no pre-selection in the dropdown"
- suggested_pending_cs = key.clin_sig — current behaviour, pre-selects the current value (no change) in the dropdown

The comment says "leave it as it was" implying pass, but None might be better UX (the dropdown shows no pre-selection rather than silently favouring the first one
found). This depends on what the discordance action UI should show when labs have conflicting pending directions.