Do not call payload.num_transactions()

[lukaszrzasik]

Closes #<ISSUE_NUMBER>

This PR:

This PR just substitutes a call to payload.num_transactions() with self.tx_queue.len() which gives the same results.
payload.num_transactions() looks deceptively harmless but the implementation in Payload is very cpu heavy.
I see big performance improvement with just this change.

This PR does not:

Unfortunately does not fix the current builder issues. Builder works visibly faster but is still very cpu-heavy and the sequencer times out waiting for the blocks.

Key places to review:

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

Lukasz Rzasik seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}