Use bounded concurrency for block lookups to prevent resource spikes on large UTxO sets.

[scarmuega]

In @crates/minibf/src/routes/utxos.rs around lines 38 - 60, The current use of
join_all on tx_deps launches unbounded concurrent calls to
domain.query().block_by_tx_hash and can spike resources; replace that pattern
with a bounded stream using futures::stream::iter(tx_deps.iter()).map/then to
build the same async task for each tx and apply .buffer_unordered(64) to limit
concurrency to 64, then .collect::<Vec<_>>().await (or continue the existing
into_iter/flatten/collect flow) to produce block_deps; keep the existing
handling logic inside the per-tx async closure (the MultiEraBlock::decode and
UtxoBlockData::try_from paths) but invoked by the buffered stream instead of
join_all so lookups for block_deps are rate-limited.