ci: Switch to reusable bench-deploy.yml

[samuelburnham]

• Adds historical benchmark plots via reusable bench-deploy.yml workflow, described in Fix bench-deploy.yml action argumentcomputer/ci-workflows#44
• Refactors Fibonacci benchmark env configuration
  • Groups benchmarks by fib_n input rather than rc, which allows for a more natural comparison for each of our output types. For example this means in gh-pages we will have one plot per Fib input and one line per RC value, which means different RC performance can be compared directly.
  • Changes the LURK_RC env var to LURK_BENCH_RC to clarify its usage in benchmarks and bench-like examples only
  • Adds the LURK_BENCH_FIBN env var to configure Fibonacci input values as desired (e.g. in a PR comment benchmark)

Note

Needs further testing once merged

[samuelburnham]

!benchmark --bench fibonacci --features cuda

[github-actions]

Benchmark for 359e6a1

Click to view benchmark

Test	Base	PR	%
Fibonacci Prove - num = 100/fib/rc-100	1457.9±9.98ms	N/A	N/A
Fibonacci Prove - num = 200/fib/rc-100	2.8±0.01s	N/A	N/A
LEM Fibonacci Prove - rc = 100/fib/num-100	1441.7±7.10ms	N/A	N/A
LEM Fibonacci Prove - rc = 100/fib/num-200	2.8±0.01s	N/A	N/A
LEM Fibonacci Prove - rc = 600/fib/num-100	1835.5±13.52ms	N/A	N/A
LEM Fibonacci Prove - rc = 600/fib/num-200	3.0±0.02s	N/A	N/A