Review mempool design as well as testing strategy, and remove any possible bottlenecks

[jules]

As the mempool is currently heavily relying on channel-based processing, it stands to reason that we can most likely increase the efficiency by a large margin.

Additionally, the mempool testing strategy should be reviewed, and possibly altered if deemed necessary. In my experience, the current test file may be a bit hard to read when you are new to it, and the singleton-based strategy (with the context object that resets a single instance of the mempool throughout the tests) has caused some regressions in the past that were purely testing-related, even though the mempool itself worked just fine. A review and a possible re-design might save us time in the future of having to catch these harmless regressions and fix them, just to have our pipeline succeed.

[autholykos]

This issue depends on phoenix integration to be done first