[frozenspider]

Resolves #41

Summary of changes:

• Change coefficients in SHA-256 circuit from BigInt to i64 - this is safe to do as all coefficients are $\le 2^{35}$.
• Relax BitzSemiring: From<u64> to From<u32>. This does not involve any kind of performance penalty, but lets i64 become BitzConstraintRing: BitzSemiring.
• As a guardrail, check_witness and IntegerProducts::is_satisfied now perform check for arithmetics overflow.
• As another guardrail, constraint generator now checks limb capacity of a ring at compile time.

Result:

Measured on my machine:
cargo run --release -p bitz-cli -- circuit-e2e --circuit sha256-2kb --threads 5

What	Before	After
Setup/Synthesize constraints	~420 ms	~300 ms
Setup/Prepare integer matrices	~290 ms	~210 ms
Peak RSS of circuit-e2e	270 MiB	186 MiB
Generator output matrices	95 MiB	54 MiB
Synthesis peak	132 MiB	89 MiB
Prepare integer matrices peak	159 MiB	89 MiB

cargo run --release -p bitz-cli -- circuit-e2e --circuit sha256-chain --num-blocks 600 --threads 5

What	Before	After
Setup/Synthesize constraints	~8.4 s	~6.9 s
Setup/Prepare integer matrices	~3.6 s	~1.3 s
Peak RSS of circuit-e2e	4790 MiB	3220 MiB
Generator output matrices	1781 MiB	1020 MiB
Synthesis peak	2328 MiB	1543 MiB
Prepare integer matrices peak	2889 MiB	1543 MiB

Other running time numbers are unaffected.