feat(#196): expose blocks/disk_growth/xvb_history on /api/state (Tier-1)

[VijitSingh97]

Summary

Partial #196 — Tier-1 persistence.

The #196 telemetry-persistence epic's backbone (PR #600, already on develop/develop-v1.11) added
five dedicated SQLite tables — blocks, xvb_history, network_history, disk_growth,
worker_history — each with its own capture hook, retention, and per-table write-health signal.
That PR was deliberately scoped to storage only: "Backbone only — capture, storage, retention, and
a getter per series; charts/UI and /api/state exposure are deliberate follow-ups."

This PR is that follow-up for the three series the parent task calls Tier-1:

• blocks — pool block-found events (permanent, event-driven)
• disk_growth — hourly monerod-DB-size + host-disk-usage samples (permanent)
• xvb_history — ~5-min XvB-credited scalar samples (30-day retention)

Each is now assembled in build_state() (build/dashboard/mining_dashboard/web/views.py) from the
existing StateManager.get_blocks() / get_disk_growth() / get_xvb_history() getters, range/window
-filtered the same way as the existing share_stats series, and bucket-averaged past the existing
700-point chart cap for disk_growth/xvb_history (both can exceed that cap over a long-lived
install; blocks can't — a handful of rows a week — so it isn't downsampled). A ponytail-review pass
folded the two gauge-series formatters into one shared helper (net -13 lines).

Deferred / out of scope (per the parent task's explicit instructions):

• network_history and worker_history (Tier-2) are not exposed here — separate slice of the epic.
• No chart/UI renders any of these three series yet. Storage, capture, retention, and now API
  exposure are done; charting is a further follow-up.
• Pool-wide share-health (share_stats) was already shipped under Store share stats as time-series for reject-rate trends/charting #116 — untouched.

Because the storage/capture/retention/counter-reset backbone (including its own tests) was already
merged in #600, this PR is scoped to exactly the missing piece: the /api/state formatters, their
wiring, and their tests.

Targets the v1.11 integration branch; merges to develop after v1.10 ships.

Test plan

• make lint — all surfaces pass except lint-proto (needs a Docker daemon, unavailable in
  this sandbox; unrelated to this change — the vendored Tari protos aren't touched).
• make test-dashboard — 1472 passed, 96.66% overall coverage (gate: 80%).
• diff-cover coverage.xml --compare-branch=origin/develop-v1.11 --fail-under=90 — 96.3% patch
  coverage (1 pre-existing-pattern uncovered defensive line, matching the identical branch in
  the existing _downsample_share_stats/_downsample_history).
• bash tests/stack/run.sh — 1541 passed, 0 failed.
• bash tests/stack/test_compose.sh — all checks passed.
• bash tests/integration/selftest.sh — 132 passed, 0 failed.
• uv run pytest tests/integration/fakes — 23 passed.
• node --test build/dashboard/tests/frontend/*.test.mjs — 222 passed.
• make lint-docs-voice / make lint-md — clean on the doc/changelog edits.

🤖 Generated with Claude Code