feat: ELECTRS_ROCKSDB_WRITE_BUFFER_SIZE

[dpc]

In testing setups large write buffer sizes lead to large db files that potentially waste resources / interfere with CIs limits.

This change adds an ability to overwrite the default behavior

[romanz]

Thanks for opening this PR!

However, when testing on a single-block regtest instance:

[2023-12-20T19:49:08.486Z INFO  electrs::server] serving Electrum RPC on 127.0.0.1:60401
[2023-12-20T19:49:08.522Z INFO  electrs::db] "./db/regtest": 0 SST files, 0 GB, 0 Grows
[2023-12-20T19:49:08.527Z INFO  electrs::index] indexing 1 blocks: [1..1]
[2023-12-20T19:49:08.570Z INFO  electrs::chain] chain updated: tip=146494c9b151c581d322916d94a170aa7a1bf3e34f6de4d4cf6405af03f8af61, height=1
[2023-12-20T19:49:08.585Z INFO  electrs::db] starting config compaction
[2023-12-20T19:49:08.586Z INFO  electrs::db] starting headers compaction
[2023-12-20T19:49:08.587Z INFO  electrs::db] starting txid compaction
[2023-12-20T19:49:08.588Z INFO  electrs::db] starting funding compaction
[2023-12-20T19:49:08.589Z INFO  electrs::db] starting spending compaction
[2023-12-20T19:49:08.591Z INFO  electrs::db] finished full compaction

It seems that the existing settings result in a not-too-large DB:

$ du -sh db/regtest/
340K	db/regtest/

[dpc]

@romanz This is interesting. The behavior might have changed maybe between rocksdb versions? It does help with the esplora electrs fork.

Anyway, closing as it doesn't seem a problem anymore.