Verification fails on proofs which merge 1 previous proof, after conversion with Side_loaded.of_proof

[mitschabaude]

impact: this would currently prevent these proofs from being accepted in transactions. for example, the mastermind example would probably not work when deployed on-chain

to reproduce, run the snarkyjs example src/examples/program.ts, and observe the bad verify: dlog_check messages

[joseandro]

@mitschabaude has this work been deprioritized?

[mitschabaude]

@joseandro Not to my knowledge! It has to be fixed at least before the HF, otherwise we won't fully support recursion in user circuits

There's this PR, which has been open for a couple of weeks, which addresses the issue:
#11698
It seems to have a couple of integration tests failing still