fix(history): entrypoint detection was first-field-wins, plus a two-tier head read

extractTranscriptEntrypoint returned the FIRST entrypoint-bearing message's
value instead of scanning for any 'cli' occurrence, so a transcript that
started under an older Claude Code build (no entrypoint field) and later
picked up a non-'cli' entrypoint on some later message was wrongly excluded
from history — the opposite of the fail-open behavior the function's own
comment claimed. Now returns 'cli' the moment any scanned message carries it,
and only falls back to a non-cli value when nothing else qualifies. Head/tail
entrypoints are merged the same way (either side being 'cli' wins).

Also restructures scanProjectDir's head read into two tiers: try 16KB first
and escalate to 128KB only when that wasn't enough, instead of reading 128KB
for every file unconditionally. Measured against a real ~/.claude/projects
tree, the unconditional-128KB version roughly quadrupled scan cost to fix a
problem only a minority of files actually have; the two-tier version cuts
bytes read by ~36% and wall time by ~17% while producing identical output.
Also fixes a fallback regression where a failed head read (e.g. EMFILE) on a
file at or under the head buffer size no longer got a shot at the tail-read
fallback, silently dropping the session from history.

test/services/unified-session-service.test.ts

@@ -303,6 +303,42 @@ describe('mergeUnifiedSessions', () => {
     expect(old!.firstPrompt).toBeUndefined();
   });
 
+  it('leaves a RESUMED session blank rather than borrowing a sibling, once its own transcript is aliased in', () => {
+    // The exact scenario COD-140's own comment lists first: a live/persisted row
+    // whose claudeSessionId aliases to an on-disk transcript. Once that alias
+    // successfully folds the transcript's own (failed) extraction into this row
+    // (sources includes 'history'), it must NOT then fall through to the
+    // workingDir guess and borrow an unrelated sibling's prompt -- same bug as
+    // the plain history-only case above, but for the resumed-session path the
+    // backfill mechanism was actually built for.
+    const merged = mergeUnifiedSessions({
+      live: [{ id: 'codeman-resumed', status: 'working', claudeSessionId: 'resumed-uuid', workingDir: '/shared' }],
+      history: [
+        // The resumed session's OWN transcript -- aliased in via claudeSessionId,
+        // but its own extraction found nothing.
+        {
+          sessionId: 'resumed-uuid',
+          workingDir: '/shared',
+          sizeBytes: 5000,
+          lastModified: '2026-01-01T00:00:00.000Z',
+          firstPrompt: undefined,
+        },
+        // An unrelated, newer sibling in the same directory.
+        {
+          sessionId: 'sibling-uuid',
+          workingDir: '/shared',
+          sizeBytes: 6000,
+          lastModified: '2026-06-01T00:00:00.000Z',
+          firstPrompt: "unrelated sibling's prompt",
+        },
+      ],
+    });
+    const resumed = merged.find((m) => m.sessionId === 'codeman-resumed');
+    expect(resumed).toBeDefined();
+    expect([...resumed!.sources].sort()).toEqual(['history', 'live']);
+    expect(resumed!.firstPrompt).toBeUndefined();
+  });
+
   // COD-145: lastPrompt backfill — mirrors the COD-140 firstPrompt path so the
   // most-recent user prompt also reaches live rows whose id ≠ transcript UUID.
   it('backfills lastPrompt onto a live session by claudeSessionId join (uuid-join)', () => {