fix: remove duplicate status:2 attempt writes in TreeSearchProbe

[Christbowel]

What this fixes

TreeSearchProbe.probe() was calling the detector internally and then
writing each attempt directly to the report file as status:2. The harness
(harnesses/base.py) independently writes the same attempts as status:2
after probe.probe() returns — this is the standard flow for all probe types.

The result: every attempt produced by TreeSearchProbe appeared once as
status:1 and twice as status:2 in the report, with identical UUIDs
and content. Verified on topic.WordnetControversial which showed 56/56
UUIDs affected.

The fix removes the 4 redundant lines from TreeSearchProbe.probe().
The detector call is kept, TreeSearchProbe needs the detector results
internally to make tree-search decisions (continue exploring or not).
Only the write is removed.

Fixes #1522

Verification

# Before fix: every UUID appears 3 times (1x status:1, 2x status:2)
python -m garak --target_type test.Blank --probes topic.WordnetControversial --generations 1

# After fix: every UUID appears exactly twice (1x status:1, 1x status:2)
python3 -c "
import json
from collections import defaultdict
counts = defaultdict(lambda: defaultdict(int))
with open('<latest_report>.report.jsonl') as f:
    for line in f:
        d = json.loads(line)
        if d.get('entry_type') == 'attempt':
            counts[d['uuid']][d['status']] += 1
dupes = {u: s for u, s in counts.items() if s.get(2, 0) > 1}
print(f'UUIDs with duplicate status:2: {len(dupes)}')  # Expected: 0
print(f'Total UUIDs: {len(counts)}')                   # Expected: 56
"

All probes inheriting from TreeSearchProbe are affected:
topic.WordnetControversial, topic.WordnetBlockedWords,
topic.WordnetAllowedWords.

[jmartin-tech]

Thanks, this makes sense.