[Refactor] Update `GraphEncoding` and `ExecutionTarget` in `wasmedge-sdk::plugin`

[apepkuss]

In this PR, refactor GraphEncoding and ExecutionTarget types to keep their names consistent with the counterparts defined in wasi-nn crate (v0.6.0).

In addition, support for the conversion from Strings to Enum variants.

[juntao]

Hello, I am a code review bot on flows.network. Here are my reviews of code commits in this PR.

---

Overall Summary:

The pull request titled "[Refactor] Update GraphEncoding and ExecutionTarget in wasmedge-sdk::plugin" introduces several changes to the codebase. The key changes include adding new variants to the GraphEncoding and ExecutionTarget enums, updating the implementations of Display and FromStr for the GraphEncoding enum, and adding a conditional import.

While most of the changes are straightforward, there are a few potential issues and errors that need to be addressed. These include a typo in the formatting of a variant in the NNBackend enum, the need for additional changes to fully support the new variant NNBackend::TensorFlow, outdated comments and code in the NNPreload struct, and incomplete error handling in the implementations of FromStr for both enums.

It is important to address these issues and errors before merging the pull request.

Details

Commit 69358fb6ea3cb18d910de3f5329f61b0ea7bde5c

Key changes:

• Added a new variant NNBackend::TensorFlow in the NNBackend enum.
• Added a new variant NNBackend::ONNX in the NNBackend enum.
• Updated the fmt implementation for NNBackend to include the new variants.

Potential problems:

• In the NNBackend enum, the variant NNBackend::TensorFlowLite is formatted as "TensorflowLite" instead of "TensorFlowLite". This might be a typo.
• The new variant NNBackend::TensorFlow might require additional changes elsewhere in the codebase to fully support it.

Overall, the changes seem to be straightforward and introduce new options for the NNBackend. However, the typo in the formatting and the potential impact of the new variant should be addressed.

Commit 78d44584244116aa56a9fd629faf030faa8473d7

Key changes:

• Added use wasmedge_types::error::WasmEdgeError to the imports.
• Renamed NNBackend enum to GraphEncoding.
• Added new enum variants: OpenVINO, ONNX, TensorFlow, Autodetect, and GGML.
• Updated the implementation of Display for GraphEncoding enum.
• Added the implementation of FromStr for GraphEncoding enum.
• Added new enum variant AUTO to ExecutionTarget.

Potential problems:

• The comment describing NNBackend in the code has not been updated to reflect the changes to the enum.
• The implementation of FromStr for GraphEncoding enum should return an error if the input string does not match any of the enum variants.
• The implementation of FromStr for ExecutionTarget enum should return an error if the input string does not match any of the enum variants.
• The NNPreload struct still uses the old NNBackend enum instead of the new GraphEncoding enum. This should be updated.

Overall, the key changes involve renaming and updating the enums NNBackend and ExecutionTarget. Some comments and code in the NNPreload struct need to be updated accordingly. Additionally, the implementations of FromStr for both enums should handle unknown input strings correctly by returning an error.

Commit 0d93da98eb0d270fc4b2a8236c693c0003056a00

Key changes:

• The import wasmedge_types::error::WasmEdgeError is conditionally used when the feature flag wasi_nn is enabled.

Potential problem:

• It seems that there is only one small change in this patch, which is adding a conditional import. Therefore, there might not be any significant issues to address in this pull request.