[PyTorch] Debug CUDA graph support with operation-based API (#1117)

* Debug CUDA graph support with operation-based API

Signed-off-by: Tim Moon <[email protected]>

* Refactoring CUDA graph tests

Signed-off-by: Tim Moon <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Review suggestions from @ptrendx

Return default recipe from FP8GlobalStateManager.get_fp8_recipe if needed. Expand error message when failing to load FP8 state after capturing CUDA graph.

Signed-off-by: Tim Moon <[email protected]>

* Avoid unnecessary recursion when saving/loading FP8 state

Signed-off-by: Tim Moon <[email protected]>

* Fix circular import

Signed-off-by: Tim Moon <[email protected]>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Signed-off-by: Tim Moon <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>