Defer payload processing to next block

[nflaig]

Under the current gloas spec, a single block root can correspond to two different post-states depending on whether the payload has been seen. This breaks several spec invariants that hold on mainnet since phase0

• Dependent root: execution requests (like consolidations) can change effective balances between the post-block and post-payload state, so the same block root no longer guarantees identical balance weighted assignments (eg. proposer_lookahead and ptc_window) across epoch boundaries
• Checkpoint state: store_target_checkpoint_state uses the post-block state while fork choice may use the post-payload state, producing divergent epoch transitions and weakening FFG guarantees (@potuz may disagree). This affects justified balances in fork choice weight calculations which is currently underspecified as noted in Fork-choice justified balances in underspecified post-Gloas #5074
• Weak subjectivity: is_within_weak_subjectivity_period and compute_weak_subjectivity_period operate on a checkpoint state that is now ambiguous due to the post-block state missing execution request effects

This PR defers execution payload processing (except verification) to the next beacon block's state_transition via a new process_parent_execution_payload function. process_execution_payload remains as a pure verification step on payload arrival. This eliminates the dual state architecture (block_states / payload_states) introduced in the current gloas spec and restores a single canonical post-state per block root, restoring all three invariants noted above without requiring more invasive changes like updating the Checkpoint type as proposed here as an alternative solution.

cc @wemeetagain, @ensi321, @twoeths

[jtraglia]

All, this PR has grown quite large (~100 comments & 60 commits). It's clear to me that (1) we agree that this is appropriate strategy and (2) it's in a good enough state to be merged. I'm sure there are little tweaks we can make, but let's merge this and follow up with smaller, more targeted PRs. Great work everyone!