docs(handover): add Indonesian scope summary + its generator

PFM-Scanner-Ringkasan-Scope.pdf/.odt — 20 pages covering the flow, folder map,
ports, how to run each part and from which directory, how login and upload move
data, and what is built versus what isn't.

src/build_ringkasan.py draws its four diagrams as native shapes (same approach as
build_deck.py), crops them to img/ring-*.png, and builds the document in one run.

docs/handover/README.md

@@ -0,0 +1,107 @@
+# Handover materials
+
+Two deliverables, same content, different jobs.
+
+| File | What it's for |
+|---|---|
+| `PFM-Scanner-Knowledge-Transfer.odp` | 30-slide deck (LibreOffice Impress). For the walkthrough session. Diagrams are native draw shapes, so it also opens in Draw. |
+| `PFM-Scanner-Handover-Document.odt` | 42-page document (LibreOffice Writer). For the person working alone, later. Adds an operational runbook, a troubleshooting guide, a glossary, and a full screenshot appendix. |
+| `PFM-Scanner-Ringkasan-Scope.odt` | 20-page Indonesian summary (LibreOffice Writer). The one to read first: the flow, the folder map, the ports, how to run each part and from which directory, how login and upload actually move data, then what is built versus what isn't. |
+| `*.pdf` | Same files, for sharing with people who won't edit them. |
+
+All three are built entirely from what is committed in this repo. Nothing is invented.
+Screenshots are the real captures from `screenshots/v2/` and `screenshots/`, taken on
+a real device against the live backend.
+
+## Before handing them over — fill these in
+
+Three things the repo cannot know are marked `[ TO FILL IN ]`:
+
+- **Deck slide 1 / document cover** — presenter name, handover date, successors.
+- **Deck slide 24 / document §15** — the access & credentials checklist is a
+  *checklist*, deliberately unfilled. Don't type secrets into either file; use it as
+  the agenda for a session where the successor logs into each one themselves.
+- **Document §12.5, §12.7, §15** — how the APK reaches store devices, whether any DB
+  backup runs, and the current deployment state (which stores are live, on what
+  hardware, since when).
+
+## Contents
+
+**Deck (30 slides)** — 1–2 intro · 3–8 what exists + architecture and sequence
+diagrams · 9–11 the app on screen (shell, DO scan ×2) · 12–13 gateway + internal web
+tooling · 14–16 accuracy · 17–18 Product Scan + its screens · 19–21 data, ops, repo
+map · 22–24 gotchas + credentials · 25–30 roadmap, scope, first 30 days, index.
+
+**Document (42 pages)** — §1–9 what exists · §10 accuracy · §11–13 setup, runbook,
+troubleshooting · §14–18 gotchas, credentials, roadmap, out-of-scope, first 30 days ·
+Appendix A glossary · B command reference · C doc index · D the app screen by screen.
+
+## Regenerating
+
+`src/` holds the generators. Both emit flat-ODF XML that LibreOffice converts.
+
+```bash
+cd docs/handover/src
+
+python3 prepare_assets.py     # once — builds shots_small/ from ../../../screenshots/
+                              # (needs Pillow; nothing else here does)
+
+python3 build_deck.py                                     # -> pfm-handover.fodp
+libreoffice --headless --convert-to odp pfm-handover.fodp
+
+python3 build_doc.py                                      # -> pfm-handover-doc.fodt
+libreoffice --headless --convert-to odt pfm-handover-doc.fodt
+```
+
+`shots_small/` is gitignored — it is display-resolution copies of screenshots already
+committed elsewhere in the repo, and embedding the originals at full size makes the
+`.odp` roughly 10MB for no visible gain.
+
+The Indonesian summary has its own generator, same two-stage shape as the deck:
+
+```bash
+cd docs/handover/src
+python3 build_ringkasan.py     # draws diagrams, crops them, builds the doc,
+                               # writes ../PFM-Scanner-Ringkasan-Scope.{pdf,odt}
+```
+
+Its four diagrams are native shapes like the deck's, drawn in the same file that
+builds the document, cropped to `img/ring-*.png`. Edit the shape coordinates in
+`build_ringkasan.py` and re-run — there is no separate re-crop step.
+
+Two layout constraints it works around, worth keeping if you edit it. Writer glues a
+heading to a full-width image below it, so a section opening with a diagram jumps to
+the next page and strands a half-empty one — `newpage()` makes that jump deliberate.
+And a sequence diagram wider than about 20cm shrinks below readable size once scaled
+into the A4 text column, so keep the crop narrow rather than the font large.
+
+`build_deck.py` prints a text-overflow report for any frame whose content is
+estimated to exceed its box — anything over ~0.15cm is worth fixing before shipping.
+
+### If you change a deck diagram
+
+The document embeds the deck's diagrams as PNGs from `src/img/`, cropped out of the
+deck PDF. Re-crop after any diagram change, or the two will disagree. The crop
+rectangles are in centimetres `(deck page, x, y, w, h)`; `build_doc.py` resolves them
+by filename prefix, so the page number in the filename can change freely.
+
+```bash
+# after rebuilding the deck PDF
+python3 - <<'PY'
+import subprocess
+DPI=200; K=DPI/2.54
+crops={"flow":(3,1.35,4.40,22.70,2.90),"arch":(4,1.30,3.10,22.90,5.05),
+       "sequence":(6,1.00,3.25,23.20,9.15),"layers":(14,1.35,4.25,22.70,7.00),
+       "product":(17,1.35,4.25,22.70,2.20),"phases":(27,1.35,8.75,22.70,2.05)}
+for n,(p,x,y,w,h) in crops.items():
+    subprocess.run(["pdftoppm","-r",str(DPI),"-f",str(p),"-l",str(p),"-png",
+        "-x",str(int(x*K)),"-y",str(int(y*K)),"-W",str(int(w*K)),"-H",str(int(h*K)),
+        "pfm-handover.pdf","img/"+n],check=True)
+PY
+```
+
+Note: LibreOffice installed from snap cannot read paths outside `$HOME`, so run the
+conversions somewhere under your home directory.
+
+Editing directly in Impress/Writer is fine, but then the generators are stale — pick
+one as the source of truth rather than maintaining both.