fix(gpu): fix compression throughput benchmark

[pdroalves]

closes: https://github.com/zama-ai/tfhe-rs-internal/issues/1166

PR content/description

Check-list:

• Tests for the changes have been added (for bug fixes / features)
• Docs have been added / updated (for bug fixes / features)
• Relevant issues are marked as resolved/closed, related issues are linked in the description
• Check for breaking changes (including serialization changes) and add them to commit message following the conventional commit specification

---

This change is 

[pdroalves]

@agnesLeroy We still need to validate the new numbers of elements for 64 and 128-bit compression. These values are good for 4xL40 but I wasn't able to launch SXM5 to see how it goes there.

[pdroalves]

@soonum This PR splits 64 and 128-bit compression benchmarks. This way we are more consistent (as we do with PBS) and it's faster to bench 128-b compression. Otherwise we have to wait forever for the 64-bit compression be benchmarked, discarded, and only then run 128-b.

[agnesLeroy]

These benches will only run on SXM5 so we need to fine tune this number for there.

[IceTDrinker]

leftover logging in tfhe-zk-pok needs to be removed

a rename would be nice to match the PR https://reviewable.io/reviews/zama-ai/tfhe-rs/2814 from David that reworks some benchmarks to be easier to use for perf regressions

@IceTDrinker reviewed 5 of 5 files at r1, all commit messages.
Reviewable status: all files reviewed, 5 unresolved discussions (waiting on @pdroalves and @soonum)

---

tfhe-zk-pok/src/curve_api/bls12_446.rs line 317 at r1 (raw file):

        #[track_caller]
        pub fn multi_mul_scalar(bases: &[Self], scalars: &[Zp]) -> G2 {
            let start = Instant::now();

needs to be removed

---

tfhe-zk-pok/src/curve_api/bls12_446.rs line 329 at r1 (raw file):

            };
            println!(
                "multi_mul_scalar: {} bases, {} scalars, {:.2} ms",

same here

---

tfhe-benchmark/Cargo.toml line 100 at r1 (raw file):

[[bench]]
name = "glwe_packing_compression_128b-integer-bench"

naming to be checked

---

Makefile line 1380 at r1 (raw file):

	RUSTFLAGS="$(RUSTFLAGS)" __TFHE_RS_BENCH_TYPE=$(BENCH_TYPE) \
	cargo $(CARGO_RS_CHECK_TOOLCHAIN) bench \
	--bench	glwe_packing_compression_128b-integer-bench \

there is a renaming in progress @soonum in your PR if you can advise on the "correct" naming here so that both merges are in accordance

[IceTDrinker]

let me know if my review is blocking here

[pdroalves]

@IceTDrinker not really. I'm still trying to find the optimal number of elements for each bit size. With the scarcity of sxm5s and another task I'm working on, it's taking some time.