feat[gpu]: jit kernels

[joseph-isaacs]

Add support for jitting bitpacking, FoR and alp kernels to cuda.

[cloudflare-workers-and-pages]

Deploying vortex-bench with    Cloudflare Pages

Latest commit:	916f93e
Status:	✅  Deploy successful!
Preview URL:	https://c62bdc1f.vortex-93b.pages.dev
Branch Preview URL:	https://ji-jit-bp-for.vortex-93b.pages.dev

View logs

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.37%. Comparing base (0eea6c8) to head (a81472f).
⚠️ Report is 31 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]

this todo made me go down the rabbit hole https://stackoverflow.com/questions/32735292/can-anyone-provide-sample-code-demonstrating-the-use-of-16-bit-floating-point-in. You need an include and then you can use half but you really want to use half2, i.e. pack two halfs into 32 bits

[robert3005]

Not quite sure how continuation passing style would look like differently? You are nesting the code emitting statements recursively at any given step. You could have a function that will in turn when called emit the code instead of directly emitting the code BUT I can't quite assess in abstract if its easier... maybe?