add preallocated and indexed MSMs

[winston-h-zhang]

Upstream fixes for argumentcomputer/grumpkin-msm#13 and argumentcomputer/arecibo#306, just switching branches and benchmarking

[winston-h-zhang]

Welp hopefully there's no perf regression

!gpu-benchmark

[github-actions]

Benchmark for 0c4c6e9

Click to view benchmark

Test	Base	PR	%
LEM Fibonacci Prove - rc = 100/fib/num-100	1459.6±10.36ms	2.5±0.01s	+71.28%
LEM Fibonacci Prove - rc = 100/fib/num-200	2.8±0.01s	4.8±0.01s	+71.43%
LEM Fibonacci Prove - rc = 600/fib/num-100	1826.5±13.73ms	2.6±0.02s	+42.35%
LEM Fibonacci Prove - rc = 600/fib/num-200	3.0±0.02s	4.3±0.02s	+43.33%

[winston-h-zhang]

Oh noes...

[winston-h-zhang]

!gpu-benchmark

Should be fixed

[github-actions]

Benchmark for bea2d02

Click to view benchmark

Test	Base	PR	%
LEM Fibonacci Prove - rc = 100/fib/num-100	1451.6±9.94ms	1466.8±5.86ms	+1.05%
LEM Fibonacci Prove - rc = 100/fib/num-200	2.8±0.02s	2.8±0.02s	0.00%
LEM Fibonacci Prove - rc = 600/fib/num-100	1854.7±3.84ms	1878.1±15.01ms	+1.26%
LEM Fibonacci Prove - rc = 600/fib/num-200	3.1±0.01s	3.1±0.02s	0.00%

[huitseeker]

Closing in favor of the argumentcomputer/arecibo#374 pipeline