apollo_staking,apollo_consensus_manager: read mock epoch height from batcher first (#14698)

[matanl-starkware]

Backport of #14698 to main-v0.14.3.

Clean git cherry-pick -x of 163e4fc103. No starknet_version change (V0_14_3 preserved).

Stacked on matanl/bp-14689-next-epoch-window — merge that first.

[reviewable-StarkWare]

This change is 

[cursor]

PR Summary

Medium Risk
Changes how consensus committee epochs are computed in the mock staking path; incorrect height would still stall block production, though behavior is narrowed with batcher-first logic and an existing state-sync fallback.

Overview
Mock staking epoch resolution now uses the batcher’s committed chain tip instead of state sync’s latest block when computing the current (and thus previous) epoch.

MockStakingContract takes a SharedBatcherClient and resolves height via latest_committed_block: get_height() minus one (next-to-build marker), with a warn and fallback to state sync on batcher failure. create_committee_provider passes the batcher client into the mock contract constructor.

Tests were rewired to mock batcher height markers and add coverage for epoch-boundary markers and batcher-error fallback. The MIN_EPOCH_LENGTH comment in staking_manager was updated to describe lag relative to the batcher height source.

^{Reviewed by Cursor Bugbot for commit 163e4fc. Bugbot is set up for automated code reviews on this repo. Configure here.}