feat(tests): limit number of nilchain payers in tests

[andreasbros]

Motivation

There is a need to limit the number of payers (blockchain payment accounts) to be created when running functional tests. Right now fixture creates 100 of them.

Solution

This change introduces a pool-based approach using ManagedNillionChainClientPayer that creates a fixed number of payer instances (limited by MAX_PAYERS_NUM) and hands them out in a round-robin fashion via an atomic counter.

If multiple tests end up sharing the same payer concurrently, the call submit_payment() is called concurrently but is guarded by an async mutex (inner), only one task acquires the lock at a time, effectively serialising the payment submissions.

Note: MAX_PAYERS_NUM is set to 2, and this doesn't affect func tests peformance locally - they finish in 12 - 20 seconds.

Fixes #
Design discussion issue (if applicable) #

Merge requirement checklist

• CONTRIBUTING guidelines followed
• Unit tests added/updated (if applicable)
• Breaking change analysis completed (if applicable). "Will this change require all network cluster operators to update? Does it break public APIs?"
• For new features or breaking changes, created a documentation issue in nillion-docs

[mfontanini]

What was the reasoning behind moving this logic to a new type? This pooling logic could have been here right? You essentially moved the existing funded_payers to a new type.

[mfontanini]

The idea here was to fund lots of payers in a single TX to speed tests up. If this doesn't slow anything down we might as well have a single payer and get rid of all this pooling logic? e.g. if going from 40+ payers to 2 doesn't slow it down then making it go down to 1 won't either.

I think this was added because of bors r+ tests though, so you likely won't see the slowdown locally. Did you see how much slower it is there?

[andreasbros]

some test(s) failed when running all of them with 1 payer, i think some tests require two payers?, but let me double check that