refactor: Use tox rng to seed the keypair generation.

[iphydf]

Fixes #2649

---

This change is 

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Comparison is base (961891d) 73.02% compared to head (5d7b7a7) 73.02%.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #2671 +/- ## ========================================== - Coverage 73.02% 73.02% -0.01% ========================================== Files 148 148 Lines 30482 30484 +2 ========================================== + Hits 22261 22262 +1 - Misses 8221 8222 +1 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[Green-Sky]

this makes key generation deterministic. Did you check how the fuzzer behaves with this?

[iphydf]

What behaviour do you want me to check regarding the fuzzer?

[Green-Sky]

Sorry, my real question was: "Is this intentional, with all its side effects" :)

[iphydf]

Yes, it is intentional to make tests and fuzzers deterministic.