[docs] Fix QuantizedTensor.from_float examples to use layout name strings

[BhagyeshKothalkar]

Summary

This PR updates the QuantizedTensor.from_float(...) examples to match the current API.

What changed

• Fixed the README example to pass the registered layout name string
• Fixed the TensorCoreFP8Layout docstring example to do the same

Why

QuantizedTensor.from_float(...) currently expects layout_cls to be the registered layout name string, not the layout class object itself. Internally, the implementation resolves that string through the LAYOUTS registry in comfy_kitchen.tensor.base. While the codebase and test suite correctly pass this string, the README examples and docstrings pass the layout class object.

As a result, passing TensorCoreFP8Layout directly raises a lookup error, while passing "TensorCoreFP8Layout" works as intended. This PR updates the examples to reflect the behavior implemented by the current API.

Scope

Documentation only; no implementation changes.

[BhagyeshKothalkar]

@maxious @guill @yoland68 @robinjhuang It is a documentation-only fix to ensure the QuantizedTensor.from_float examples match the current API behavior and avoid lookup errors.

The PR is ready, but it appears to be waiting on a workflow approval from a maintainer. Could someone please take a look when you have a moment? Also, please let me know if there is anything else you need from my side to get this merged. Thanks!