log: surface BTC verify backend and Esplora fallback

[LandynDev]

What

Adds debug logging so the BTC verification path makes its backend choice visible: whether a lookup was served by the local RPC node or fell back to Esplora.

Why

In node mode the validator's local bitcoind is used purely for verification reads (getrawtransaction by hash). A pruned node (our docker-compose.vali.yml default) only resolves mempool/wallet txs, so once a tx confirms and leaves the mempool the RPC returns error -5 and we silently fall back to Esplora — on essentially every verification. That reliance was invisible in logs, and a contradictory tx ... not found debug fired even when the RPC had actually errored.

(txindex=1 would fix the RPC side but is incompatible with -prune and implies a full ~700GB archive node — out of scope; we stay pruned and make the Esplora reliance observable instead.)

Changes (logging-only, behavior-neutral)

• fetch_matching_tx: log which backend served a verification (local RPC vs Esplora fallback), so reliance is greppable instead of silent.
• Short-circuit lightweight mode straight to Esplora, skipping a no-op RPC call (and the misleading log it produced).
• Drop the misleading tx ... not found debug — a None from RPC can mean "absent" or "RPC error" (already logged by rpc_call); the caller now owns the fallback log.
• get_balance: same fallback debug + skip the no-op RPC call in lightweight mode.

No control-flow or security behavior changes.

Test

• ruff format / ruff check: clean
• pytest tests/test_bitcoin_signing.py: 33 passed