Fix MemoryError during HTML report generation (port of iLEAPP #1746)

[abrignoni]

What this does

Ports the iLEAPP MemoryError fix (abrignoni/iLEAPP#1746, fixed by JSup in iLEAPP commit 0a4aee10) to RLEAPP. Report generation used to read each artifact's entire .temphtml page into memory and rebuild it with string concatenation to inject the navigation sidebar. On large returns an artifact page can reach several GB, so that peaked at roughly three times the page size and raised MemoryError.

The new stream_insert_sidebar_code() copies each page to its final location in fixed size chunks. The sidebar placeholder sits near the top of every page, so only a small head buffer is held while searching for it (including a placeholder split across a chunk boundary). The large table body that follows is streamed with shutil.copyfileobj. Output is byte identical to the old approach.

Verification

• Ran the iLEAPP regression checks locally against this port: byte identical output across page shapes (placeholder at start, end, split across a chunk boundary, missing, empty file), and peak memory bounded under half the file size on a 16 MB+ synthetic page. Not committed as a test file because iLEAPP removed its copy from main after the fix merged; the port matches iLEAPP main as it stands.
• pylint --disable=C,R on scripts/report.py: 10.00/10, zero warnings. admin/scripts/lint_changed.py against the merge base: PASS, no new warnings.
• End to end: two artifact profile run (Discord Returns friendships and server metadata) against a fully synthetic return input, since no real returns corpus is on hand. Report generated clean: sidebar navigation present in every page, zero leftover placeholders, zero .temphtml remnants, no parser errors.

To be clear about the validation boundary: the streaming path itself is exercised by the synthetic e2e run and the local regression checks. No multi GB real return was run against this port.

🤖 Generated with Claude Code