feat(fill,consume): enable `xfail`/`skip` marking of tests on a per client basis

> [!NOTE]
> **Update (2026-05-05):** I now think this is the wrong approach and we should potentially allow clients to provide a list of skip/xfails to hive. Grabbing them in Hive runs is going to be clunky, but forcing client teams to PR xfails and skips to the current dev branch and then propagating them to the relevant devnet branches will be a pain.
> Leaving open for now - the problem remains; we just need a viable solution for this.

Tldr: Add the ability for clients to mark a test case to conditionally fail or simply be skipped.

End goal:
- Single source of truth for xfails and skips for all clients (by applying pytest marks directly to `./tests`? this seems most robust?)
- Ability to add a reason for the xfail/skip.
- Propagation of xfails/skips with reason to hiveview (and other) results for a transparent overview (add registered skips/xfails to the `info` section of the test fixture JSON?).

Two example use cases (Besu):

1. Besu's client no longer supports reorgs post-Merge (as its handled by the CL), so can't be expected to pass `UncleFromSideChain_Cancun`. The test is valid for clients that do maintain this functionality, so shouldn't necessarily be removed.
2. Besu via `consume-rlp` fails, but passes `consume-engine` with `tests/cancun/eip4844_blobs/test_excess_blob_gas_fork_transition.py::test_invalid_pre_fork_block_with_blob_fields[fork_ShanghaiToCancunAtTime15k-blockchain_test-excess_blob_gas_present_False-blob_gas_used_present_True]` (and variations). This is due to the RLPBlockImporter not entirely respecting the Engine API validation rules and would require some, other unnecessary, work to fix.

Requires ethereum/execution-specs#1537.