fix(esplora): treat parallel_requests of 0 as 1

fix(esplora): treat parallel_requests of 0 as 1

A `parallel_requests` of 0 made `full_scan` and `sync` skip every request
and return an empty `Ok`, because each fetch loop takes its next batch with
`.take(parallel_requests)` and stops when the batch is empty.

Clamp it to 1 in each fetch loop, for both the blocking and async clients,
as `stop_gap` already is. Document this on `full_scan` and `sync`, and add
tests covering the spk, txid and outpoint paths.

Fixes #2291