Sync supported assets to dataset pages and refresh docs from the API

[AlexCatarino]

Summary

Two generator changes that keep the dataset pages, the docs, and the repo's generated tables from drifting apart.

sync_supported_assets.py pushes the supported-securities tables to the dataset pages. Resources/datasets/supported-securities/ is regenerated weekly from the symbol-properties database, but the same tables are the "Supported Assets" section of a dozen dataset listings, where they were pasted by hand. They had fallen behind — Binance listed 3,320 pairs against 3,651 generated ones. The script renders each section from the repo files and writes it through market/sections, keeping the Introduction sentence, the Data Summary "Asset Coverage" row and the sidebar reach in step by substituting the number rather than the sentence. It only overwrites sections it maps, never creates one, and re-reads to confirm every write. Its workflow reports on a PR and pushes on a merge to master, so the weekly generator PR shows the diff before it reaches the live site.

The alternative-dataset generators no longer take page content from the S3 dump. They were built entirely from alternative-data-dump-v2024-01-02.json, which only changes when someone re-uploads it, so listing edits never reached the docs. get_dataset_listings() splits the two jobs the dump was doing: it still says which datasets are public, because nothing in the API records that, while content comes from market/sections/read. Drift between the two sources is printed rather than acted on.

code_generators.yml gains QUANTCONNECT_ORGANIZATION_ID, which market/sections/read requires on every call.

Test plan

• sync_supported_assets.py --dry-run against the live pages — reports 13 in sync, 1 held
• Stale path exercised against a doctored copy of the tables: ticker delta, count rewrite in all three places, and the reach value
• Write-and-verify path exercised for real with an identical-bytes payload — update → re-read → diff all run, page unchanged
• get_dataset_listings() diffed against the dump: of the listings in both, every section byte-identical except the pages synced this week
• Alternative-Datasets-Code-Generator.py — 16 files changed (the synced pages' content), no additions or deletions, no renumbering
• Alternative-Datasets-Skill-Generator.py — no diff
• skill-templates/bundle-skills.py — builds all 33 skills

Generated output is not included; it lands through the weekly "Update auto-generated code" PR as usual.

Notes

algoseek-us-future-options is deliberately held out of the sync: futureoption/supported-contracts.html reads "(16) Futures Options" above 15 items, because future-table-code-generator.py counts len(FUTURE_OPTIONS) while DC is dropped — 'Dairy' is missing from the category list in get_data(). The live "(15)" is correct, so nothing is pushed until the generator is fixed.