Fix zero conflict misses when using mixed locations

[sryap]

Summary: Before this diff, TBE would assume that all rows were placed in cache and only load data from cache if the number of unique conflict misses during cache lookup was zero. However, this assumption is not accurate when the embedding locations are mixed, with some tables on HBM/UVM and others in cache. To address this issue, this diff makes TBE check the placement of each table before deciding whether to exclusively load data from cache.

This diff also refactors the forward kernel code by moving the common code into the main kernel.

Differential Revision: D54237732

[facebook-github-bot]

This pull request was exported from Phabricator. Differential Revision: D54237732

[netlify[bot]]

✅ Deploy Preview for pytorch-fbgemm-docs ready!

Name	Link
🔨 Latest commit	3092a81cf250605897b434e7b86556aa451ebf1a
🔍 Latest deploy log	https://app.netlify.com/sites/pytorch-fbgemm-docs/deploys/65dd9dc4339d0900085ed221
😎 Deploy Preview	https://deploy-preview-2360--pytorch-fbgemm-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[facebook-github-bot]

This pull request was exported from Phabricator. Differential Revision: D54237732