Galexie: Optimize resumeable start range based on core's LCL

[chowbao]

this feature was discussed here on pr for resumability.

Paraphrasing the requirements from discussion:

if we have a buckets db from the previous run we should query it to find the last closed ledger because it's very likely that the last uploaded ledger is close to the LCL.

We should adjust the start of range to LCL - (bucketSize *2) to narrow down the effective size of range.

Currently, there are ~50 million ledgers in full history. To find the latest ledger in the data lake, it would take log_2(50 * 10^6) which comes out to about 25 requests to GCS. If we use the LCL from captive core, we could potentially reduce the search space to ~ 100s of ledgers or less . log_2(100) comes out to ~ 7 requests to GCS.