fix(specs/withdrawals): Sherlock updates

ops: Fix the CI builder docker context

op-program: Hint reader recovery

Docker context was set to the `Dockerfile` rather than the containing directory.

ops: Move ci-builder into GCR

Adds a CCI job to publish `ci-builder` into GCR rather than using changesets. Subsequent PRs will update the Dockerfiles across the monorepo to use the GCR-based `ci-builder` rather than the one on Docker hub.

To release a new version of the CI builder, push a tag that starts with `ci-builder/v`.

Bring more in line with the code.

op-program: Populate preimage store

Attempt to recover the hint read stream when the NextHint callback fails

op-program now uses an in-memory key-value store as the backing for all data with the same key-based retrieval and deserialization as in non-fetching mode.

fix: error log content in start.go

op-e2e: fix TestSystemMockP2P - wait for peers to mesh

CI: fix how slither determines what to detect and when to fail.

op-program: Extract stub oracles to a reusable test package

op-chain-ops: prevent invalid decoding errors

Add a HeaderRLP method to BlockInfo

Changes sources.EthClient to reference data via the BLockInfo interface.
The actual implementation is a small wrapper over a Geth header to cache the block hash.

op-node: fix Gossip mesh parameter CLI overrides [sherlock issue fix]

op-node: L1BlockInfo MarshalBinary&UnmarshalBinary migrated to writer…

The incorrect shift in assembly detector is just broken. Where possible we should otherwise prefer
disabling detectors based on severity (currently only detecting 'high' or via triage mode so that
we can keep detecting other possibly dangerous instances of that detector.

This change implements the information provided in these slither
issues:
https://github.com/crytic/slither/issues/1408
https://github.com/crytic/slither-action/issues/38

Apparently 'fail_pedantic' defaults to true, which overrides the
'exclude_*' settings and causes failures on optimization and
informational severity findings.

As is this config should:
1. Fail only on High severity findings
2. Print Medium severity findings but not fail on them.

This commit also deletes the findings in the slither db which
were less than high severity. They are now simply ignored via this
corrected config.

specs: initial fault proof program/VM specs