feat: persist and backfill historical epoch data

[zeroklaw]

Summary

Prevent historical epoch data loss by persisting raw MiningEntry data in a permanent KV key (epochs-historical), ensuring all historical ZKC prices are backfilled over time.

Problem

The Boundless upstream API only returns the most recent 100 epoch entries. As new epochs complete, older ones silently drop off the list. With the current implementation:

• Only 100 entries are ever visible
• Epochs that drop off the window are permanently lost
• 51 epochs (20–70) have zkc_price_usd: 0 because prices were never fetched for those dates

Changes

• Add HISTORICAL_CACHE_KEY (epochs-historical) — permanent KV store with no TTL
• Add mergeHistoricalEntries() — deduplicates fresh upstream data with historical entries by epoch number (fresh data wins for conflicting epochs)
• Modify onRequestGet — on every cache miss:
  1. Merge fresh upstream entries with the historical store
  2. Persist the merged dataset back to epochs-historical (no TTL)
  3. Pass the full merged dataset through buildPriceMap and normaliseEntries
• ZKC price backfill happens automatically — since buildPriceMap only fetches missing dates, feeding it the full historical dataset triggers price lookups for all epochs that were previously skipped

Architecture

Upstream (mining) → fresh 100 entries
                      ↓
              merge with KV "epochs-historical" (permanent)
                      ↓
              buildPriceMap (only fetches missing dates)
                      ↓
              normaliseEntries → KV "epochs" (2h TTL, hot cache)
                      ↓
              applyLimit per request → response

Closes #37
Closes #22 (100 epochs target superseded by "all available epochs")

[github-actions]

Preview deployment

	URL
Latest commit	https://fe261f3a.boundless-profit-explorer.pages.dev
Branch alias	https://head.boundless-profit-explorer.pages.dev

The branch alias URL always points to the latest commit on this PR.