fix: halve chunk size when suggested get_logs range still exceeds max results

[emmajam]

Problem

The get_logs_with_retry pagination logic fails when the server's suggested retry range itself exceeds the max results limit.

For example, with ~889 logs/block, the server suggests retrying with a 26-block range (~23k logs), but the 20k limit means that range also fails. On the recursive retry, the code detects that chunk_size >= original_len and gives up entirely, surfacing the error to the test runner.

This caused RPC test failures like:

get_logs: ❌ rpc1: ErrorResp(ErrorPayload { code: -32602, message: "query exceeds max results 20000, retry with the range 24476391-24476416", data: None })

Fix

Instead of giving up when the suggested chunk covers the entire (sub-)range, halve the chunk size and retry. This lets the pagination converge to a working chunk size regardless of how accurate the server's suggestion is.

Example

• Original range: blocks 24476391..=24476422 (32 blocks)
• Server suggests: retry with 24476391-24476416 (26 blocks) — still >20k logs
• Before: gives up ❌
• After: halves to 13 blocks (~11.5k logs), succeeds ✅