Doc improvements

# bdk_chain
- [ ] `CheckPoint`'s doc comment could mention it's guaranteed to always have at least one element, since it's an invariant relied upon across the project.
- [x] `merge_chains`'s doc comment is outdated.
- [x] The doc comment for the rusqlite_impl module is a placeholder: //! Module for stuff.
- [ ] Like walk_ancestors, `walk_descendants`'s doc could mention it excludes the root tx from the iterator.
- [ ] In `SpkIterator::next()`, derived_descriptor's descriptor is unwrapped because "the descriptor cannot need hardened derivation". It'd be better to have "we check it's never hardened in the constructor" as it's a stronger invariant. (Mentioning as I initially missed it was also checked in `SpkIterator::new_with_range()` by re-assigning to end the minimum between end and `BIP32_MAX_INDEX`.)

# bdk_esplora
- [ ] Top-level documentation states "A stop_gap of 0 will be treated as a stop_gap of 1" but it's not the case. After making parallel_requests requests in the first iteration of the loop, even if all indexes were detected as active, gap_limit_reached would be set to true if stop_gap is set to 0. Not only would it not be the case if stop_gap was set to 1, but it's also surprising to stop whereas there was no gap.