Add RPC provider failover for block ingestor

[dimitrovmaksim]

Resolves #6213

When the block ingestor's polling fails and the current RPC provider is unreachable, the ingestor now automatically switches to a healthy alternative provider. If no alternative providers are configured, the current one will be retried indefinetelly, similar to the current logic.

How it works

1. On do_poll() failure, the current provider is probed via eth_blockNumber first — do_poll() can fail for non-RPC reasons (DB errors, chain reorgs), so switching would not help in those cases.
2. If the current provider is unreachable, all other validated providers are probed in parallel. The first to respond is selected.
3. If all providers are unreachable, the ingestor stays on the current provider and re-probes on the next failure.
4. There is no automatic return to the original provider — the ingestor stays on whatever provider it switched to until that one fails.

Other changes

• latest_block_ptr retries are now limited to ENV_VARS.request_retries instead of infinite, so failures surface to the failover logic.
• retry_strategy uses saturating_sub to avoid underflow when limit is 0.
• Added is_reachable() to the EthereumAdapter trait for lightweight provider health checks.
• Added all_cheapest() to EthereumNetworkAdapters to expose all validated providers.