[Plonky3 Codegen] Improve the usage of `build_aux_trace_with_miden_vm`

In #515, the user can choose to delegate the aux trace generation to the miden-vm repo through the helper `build_aux_trace_with_miden_vm`.

Currently, the function takes as arguments the main trace, the challenges, and the module to generate the trace for (`MidenModule` with variants `Chiplets`, `Decoder`, `Stack`, `Range`).

However, we could improve this functionin several ways:
- we could directly build the whole aux trace at once (through the following code: https://github.com/0xMiden/miden-vm/blob/926a0c6b714cc69a52596aefad3c77b294750ddc/processor/src/trace/mod.rs#L267)
- we could have more arguments, for instance the chiplets aux builder may need additional info such as `kernel`, `ace_hints` and `final_transcript_state`
- we could optimize the conversion back from miden_vm trace to plonky3 trace: see comment _Originally posted by @adr1anh in https://github.com/0xMiden/air-script/pull/515#discussion_r2559789464_
- finally, depending on how it will be used in MidenVM (once an E2E workflow has been established we'll see what is easiest to work with)

Examine the `build_aux_trace_with_miden_vm` function and its usage in the codebase. Review the linked MidenVM trace generation code and the PR discussion for optimization ideas. Understand the MidenModule variants and what additional data each needs. The goal is to refactor the function to build the full aux trace, accept more arguments, and improve trace conversion.

Rédigé par le modèle d'indexation à partir du texte de l'issue.