refactor: use setinterval instead of while loop in tx broadcaster

[hugocaillard]

Context

It was discovered that the tx pace slows down in time.
This PR attempts to make sure that the timing is more linear and closer to the desired interval.
With this setInterval, we boradcast tx every N seconds, and it's not impacted by networks.

With the while (true), we would wait for the previous tx to be broadcasted + N seconds to broadcast the next tx.

We'll have to keep an eye on the logs, but this could ensure stable tx pace

[hstove]

I'm running this locally to make sure it works (I'm sure it will), but I also think this won't help if the issue is due to an increase in RPC latency when fetching account nonces or even when broadcasting the tx. I'm still ok with merging it, but we should monitor and see if this fixes the issue in a long-running environment.

[hugocaillard]

@hstove

this won't help if the issue is due to an increase in RPC latency

With this PR we avoid the cascading slow downs, where previous calls would delay next ones.
This ensure that we at least try to broadcast the txs at a steady pace