fix(cli): wait for deployed programs before running tests

[raushan728]

Fixes #3624

anchor test was checking validator readiness by polling get_latest_blockhash(). This returns as soon as the RPC server accepts connections well before BPF programs are actually deployed on-chain. Tests would fire immediately after and hit "Program is not deployed".

startup_wait in Anchor.toml also had no real effect since the loop broke early regardless of the configured value.

Changes:

• Extracted the readiness loop into wait_for_validator_ready()
• Added a second check: polls get_multiple_accounts() until all deployed programs return an executable account
• Program pubkeys are extracted from the --bpf-program / --upgradeable-program flags passed to the validator, so [programs.localnet] address overrides in Anchor.toml are respected correctly
• Both checks share a single Instant-based timeout from startup_wait
• Increased STARTUP_WAIT default from 5000ms to 30000ms to account for program deployment time on top of RPC liveness
• ms_wait changed to u64 negative startup_wait values are now rejected with an error at the call site via u64::try_from
• Updated error message to reflect actual failure reason

[vercel]

@raushan728 is attempting to deploy a commit to the OtterSec Team on Vercel.

A member of the Team first needs to authorize it.

[0x4ka5h]

seems like negative startup_wait values now turn into an effectively infinite wait in the new readiness helper.