perf[gpu]: fused AoT for and bitpacking kernel

[joseph-isaacs]

I have only implemented a PoC fused FoR-BP kernel, I don't want to implemented them all since there will be a lot of duplication. I think we likely need to compile these at runtime.

I have also fixed up the kernels build system.

Fused is fast

gpu_for_bp_fused_decompress_kernel_only/u32/1GB
                        time:   [5.5376 ms 5.5410 ms 5.5443 ms]
                        thrpt:  [180.37 GiB/s 180.47 GiB/s 180.58 GiB/s]

Signed-off-by: Joe Isaacs joe.isaacs@live.co.uk

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.63%. Comparing base (b537c67) to head (7dc9fe8).
⚠️ Report is 3 commits behind head on develop.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[robert3005]

you don't need this