Fix insertion of big block

[tdroxler]

When we insert a big block, we create too many bind variable for postgres' prepared statement.
The issue is shown in that branch: feature/insert-big-block
just run: app/testOnly org.alephium.explorer.persistence.dao.BlockDaoSpec -- -z "insert big block"

[simerplaha]

Hey @tdroxler, just like how you found, Postgres allows 32767 parameters per query.

This commit fixes this by adding the ability to split a query into 2 or more queries when parameter size grows over 32767 (configurable).

Eg: for queries with one parameter per row INSERT INTO TABLE values (?)

• if parameter count is 32767 + 1, first query gets build with 32767 parameters and second one gets 1.
  • Maybe it's more efficient to split this in two queries of equal sizes (32767 + 1) / 2. This need benchmarking.
• if parameter count is 32767 * 7, seven queries get build with 32767 parameters each.

Alternatively a simpler approach would be to revert back to how Slick's typed queries did batch insert i.e. single row per insert.

BEGIN;
INSERT INTO TABLE values (?, ?, ?, ...);
INSERT INTO TABLE values (?, ?, ?, ...);
COMMIT;

I can run benchmarks to see if there is a performance difference. What do you think?

[tdroxler]

The error was: Tried to send an out-of-range integer as a 2-byte value and I end up on that issue where we can see that actually Short.MaxValue = 32767

Let's start with your solution and as we discussed, if we feel at some point that we need some optimization we can run the various benchmarks.