Consider recording transaction delay sources to the transaction execution metadata

[nagisa]

Today each transaction we execute contains some protocol-agnostic profile metadata field that describes reasons for which the gas was burnt during the processing of the transaction.

This seems like a perfect place to also record the instances where the transaction has been delayed from executing: whether the transaction has been waiting to be included into a block, placed in the delayed receipt queue, waiting for yield/resume, or whatever other reason. It could be measured in numbers of blocks or somesuch, making the measure deterministic across validators.

Once we track this information, we'd also be able to make a prometheus metric out of this.

Does this idea make any sense? cc @bowenwang1996