reth-bench: import/export block ranges + big blocks

[mediocregopher]

Describe the feature

• reth-bench should be modified so that it can export a range of payloads to a file, and then to use an exported range for new-payload-fcu calls.

• reth-bench should be modified so that when exporting a range of blocks, it instead merges blocks into Big Blocks, ie it exports 1 block for every N real blocks, with the 1 block being composed of the txs of the N real blocks.

• reth-bench-compare should be modified so that it can also use an exported range for its reth-bench calls.

Additional context

Prior work:

An existing branch exists which modifies reth-bench to force reorgs: https://github.com/paradigmxyz/reth/tree/mediocregopher/reth-bench-reorgs

It accomplishes this by building new blocks from the txs of existing blocks. Something like this could be done to accomplish building the Big Blocks.

[Animesh-Parashar]

After looking at how reth-bench currently fetches real blocks over RPC and replays them via Engine API calls, it seems possible to support this feature without introducing synthetic block semantics.

A potential approach would be:

• Export a block range by extracting and serializing the execution-ready payloads or transaction lists, optionally grouped by N consecutive blocks.
• Treat the export strictly as a bench artifact (similar to existing CSV outputs), decoupling data acquisition from execution.
• Allow reth-bench / reth-bench-compare to consume these exported payload batches directly for newPayload / forkchoiceUpdated benchmarks.

This would preserve existing execution paths and avoid fabricating headers, base fee rules, or other consensus fields, while still enabling deterministic reuse of representative workloads.

Wanted to confirm whether this payload-centric, bench-only approach aligns with the intended direction, or if constructing fully synthetic blocks is a hard requirement.