As far as I know, no one has ever used this, and it would take
extra work to support it with the new OUTPOINT_TO_UTXO_ENTRY
table (which greatly improves performance).  Just remove the
option for now.

This code is unused, now that we have the newer (and much nicer)
pointer-handling code on lines 190-194.

Replace the OUTPOINT_TO_TXOUT, OUTPOINT_TO_SAT_RANGES, and
SATPOINT_TO_SEQUENCE_NUMBER tables with a new OUTPOINT_TO_UTXO_ENTRY
table.  This saves space by avoiding redundant key storage, and
makes updates quicker since we can write all of an outpoint's
data at once.

It also simplifies caching.  We used to have utxo_cache for
OUTPOINT_TO_TXOUT and range_cache for OUTPOINT_TO_SAT_RANGES,
while SATPOINT_TO_SEQUENCE_NUMBER was uncached; now, there's
a single unified utxo_cache which covers all three.  Additionally,
since SCRIPT_PUBKEY_TO_OUTPOINT and SEQUENCE_NUMBER_TO_SATPOINT
are storing data which is already included in the cache (just
with the keys and values swapped), we get to cache those tables
too for free.

The combined effect is that an index with --index-sats and
--index-addresses is now about 54 GB smaller and twice as fast
to build.

With the unified OUTPOINT_TO_UTXO_ENTRY table, these stats aren't
useful; they're either identical to the stats in the regular
"Committing at block height ..." message (if you're using --index-sats),
or zero (if you're not).

…table-integration