feat(fusion_helpdesk_central): findings-first wizard, explicit Generate button

The old flow fired OpenAI on wizard open with just ticket + chatter,
so the AI summary was just a paraphrase of what the user originally
reported — your engineering analysis (scope, limitations, recommended
approach) never made it to the owner. Restructure to a two-step flow:

  1. Open wizard → empty findings + empty summary, NO OpenAI call
  2. You write findings: scope / effort / approach / risk
  3. Click 'Generate Summary from Findings' → OpenAI runs with
     ticket + chatter + findings, where the prompt explicitly tells
     the model to weight findings MORE THAN the original report
  4. Review/edit, then Send

Bulk wizard mirrors the flow per line: each row gets its own
findings + summary, one 'Generate All Summaries' button fans out
parallel OpenAI calls using each line's own findings.

Updated SUMMARY_PROMPT to:
- Tell the model the support engineer's findings are authoritative
- Emit a bullet structure that leads with the recommendation, not
  the user's restated ask
- Side with findings over the original report when they conflict

New tests cover:
- default_get does NOT fire OpenAI (regression guard for auto-AI)
- Findings text actually reaches the OpenAI prompt
- Send works with a manually-typed summary (no AI in the loop)
- Existing bulk + validation paths still pass with the new shape

Also folds in the deferred code-review #7: ThreadPoolExecutor now
explicitly cancels pending futures on timeout via
shutdown(wait=False, cancel_futures=True) so a slow OpenAI day can't
hold the wizard open for ceil(N/workers)*15s.

Bumps fusion_helpdesk_central to 19.0.2.3.0.

Smoke-tested live on nexa: opening the wizard makes zero OpenAI calls;
clicking Generate with findings='My findings: scope is XL, ~8h' makes
exactly one call and the findings text is verifiably in the prompt
body received by call_openai_chat.

fusion_helpdesk_central/utils.py

@@ -23,22 +23,35 @@ _logger = logging.getLogger(__name__)
 SUMMARY_PROMPT = """You are summarising a customer support ticket for a busy executive
 who needs to decide whether to approve the work.
 
+The support engineer has reviewed this ticket and written their findings
+below (scope, limitations, recommended approach, anything the original
+reporter wouldn't have known). Treat those findings as authoritative —
+they reflect the actual engineering reality, not just the user's
+description of the problem. If the findings contradict the original
+report, side with the findings and call out the gap.
+
 Output rules:
 - 4-6 short bullet points, plain text (no markdown).
-- First bullet: the ask, in one sentence.
-- Second bullet: the business impact if approved.
-- Third bullet: the business impact if NOT approved (or "none material").
-- Optional bullets: cost / effort signals if any are mentioned.
+- First bullet: the ask, in one sentence, framed in the support
+  engineer's terms (not just a paraphrase of the original report).
+- Second bullet: the support engineer's recommendation, if they made one.
+- Third bullet: business impact if approved.
+- Fourth bullet: business impact if NOT approved (or "none material").
+- Optional bullets: scope / effort / risk signals from the findings.
 - Final bullet: open questions the approver should think about.
-- Do not invent facts. If the thread doesn't say, write "not stated".
+- Do not invent facts. If the findings or thread don't say, write
+  "not stated".
 - No greetings, no sign-offs, no preamble.
 
 Ticket title: {name}
-Original report:
+Original report from the user:
 {description_plain}
 
-Replies so far:
+Replies in the ticket thread:
 {messages_plain}
+
+Support engineer's findings (your most important input):
+{findings_plain}
 """
 
 # Bound the prompt sent to OpenAI so a runaway thread doesn't blow cost or
@@ -47,13 +60,17 @@ Replies so far:
 OPENAI_PROMPT_MAX_CHARS = 8000
 
 
-def build_summary_prompt(ticket_name, description_plain, messages):
-    """Render SUMMARY_PROMPT with ticket + chatter content.
+def build_summary_prompt(ticket_name, description_plain, messages,
+                         findings=''):
+    """Render SUMMARY_PROMPT with ticket + chatter + support findings.
 
     `messages` is a list of dicts with at minimum {author, date, body_plain}.
-    We render one line per message so the model can attribute who said what.
-    Empty inputs collapse to "(none)" so the prompt never has bare blanks
-    that the model could interpret as a missing section.
+    `findings` is the support engineer's free-text notes from the wizard —
+    their scope/effort/recommendation that the AI should weight more
+    heavily than the original user description.
+
+    Empty inputs collapse to explicit markers so the prompt never has
+    bare blanks the model could misread as missing sections.
     """
     name = (ticket_name or '').strip() or '(untitled)'
     desc = (description_plain or '').strip() or '(no description)'
@@ -67,8 +84,13 @@ def build_summary_prompt(ticket_name, description_plain, messages):
         msgs = '\n---\n'.join(lines)
     else:
         msgs = '(no replies yet)'
+    findings_text = (findings or '').strip() or (
+        '(none provided — base the summary on the user\'s report and '
+        'thread alone)'
+    )
     return SUMMARY_PROMPT.format(
         name=name, description_plain=desc, messages_plain=msgs,
+        findings_plain=findings_text,
     )