Fixes #1729 and #1736

Description

Issue in #1729:

The original code had a data race between:

• READ: where we read witReqs to create wrapperReq := &ethWitRequest{Request: ethReqShim, witReqs: witReqs}.
• WRITE: where doWitnessRequest() does *witReqs = append(*witReqs, witReq)

The race occurred because buildWitnessRequests was called asynchronously in a goroutine, allowing the main goroutine to read witReqs while the background goroutine was still populating it via append operations.

By calling buildWitnessRequests synchronously, we ensure all initial witness requests are built and appended to witReqs BEFORE the main goroutine reads the slice, eliminating the possibility of concurrent read/write access.

Issue in #1736:

The original code had a data race between:

• READ: where we read announce.time for logging/timestamping
• WRITE: handleWitnessFetchFailureExt() where it does state.announce.time = time.Now().Add(fetchTimeout)

The race occurred because this read was unprotected while handleWitnessFetchFailureExt() was writing to the same announce.time field under mutex protection. Both operations access the same memory location but with different synchronisation. By protecting this read with the same mutex used by the writer, we ensure synchronised access to announce. time, eliminating the data race.