Standardize tracing crate + implement micro-benchmarking

[winston-h-zhang]

This PR is the next-in-line from downstream PR from lurk-lab/lurk-rs#627, implementing an unified tracing ecosystem. Additionally, we add the tracing_texray crate, with allows us to easily visualize spans and benchmarks of lurk. Texray can be enabled wherever and tracing offers a lot of flexibility; please see the documentation for more info (tracing-texray).

With this, we can get nice mirco-benchmarks like this:

    RecursiveSNARK::prove_step                   219ms               ├────────────┤
      <MultiFrame as StepCircuit>::synthesize     38ms                ├─┤
      <_ as Group>::vartime_multiscalar_mul       50ms                  ├─┤
      NIFS::prove                                106ms                     ├─────┤
          AZ_1, BZ_1, CZ_1                        31ms                     ├┤
          AZ_2, BZ_2, CZ_2                        29ms                       ├┤
          cross terms                              5ms                         ┆
          T                                      716μs                          ┆
          <_ as Group>::vartime_multiscalar_mul   37ms                          ├┤
      <_ as Group>::vartime_multiscalar_mul        4ms                             ┆

Looking here suggests that we are spending a surprising amount of time computing [A-C] Z_{1, 2}... 🤔

[huitseeker]

The beginning of the comment ("forgive the horror") is not super informative especially when things aren't that horrific