bench: Poseidon2 with user supplied number of bytes

[codeblooded1729]

Main objective of the PR is to get an estimate of how expensive flat hashing of tapes would be through guest program.
to this end, we generate a vector of size input_len (deterministic, since we don't want the rng to interfere in benchmark), and use poseidon2 ecall to hash it. To find the relative time taken by the hashing part, we bench it against same code, preserving the generation of vector and loading it into memory, with hashing part excluded. The results are given below.

Few trace sizes of poseidon2 table, for different input_len are given below. (Note that trace_len = (input_len/RATE).next_power_of_two() indeed holds, as a sanity check)

input_len	trace_len
1000	128
10000	2048
100000	16384

[matthiasgoergens]

What's the status of this one? @codeblooded1729