Fix IncorrectAttestationSource error in attn. simulator

[paulhauner]

Issue Addressed

Resolves #5046

Proposed Changes

Introduces both of the fixes suggested in #5046:

1. Handles the IncorrectAttestationSource error so that a missed source attestation can be identified.
2. Doesn't try to simulate attestations when the head block is more than 2 epochs older than the wall-clock slot.

My reasoning for (2) is twofold:

1. To prevent the simulator from trying to compute future, uncached shufflings from old states while syncing. Although we're careful with this regard when verifying simulated attestations, we're not so careful when producing them.
2. To avoid bad results when syncing. This is a bit philosophical, but I intended to use this feature to gauge the performance of different hardware. Getting bad "benchmarks" during sync wouldn't help my cause. I'm open to alternative views, though!

Additional Info

NA

[michaelsproul]

V nice

I think not producing attestations during sync is fine. It could be an issue if Lighthouse were crashing or de-synchronising regularly and we wanted to see that reflected in the attestation simulation metrics, but I also think it's OK to consider that out of scope. We would probably notice such an issue looking at other metrics/logs.