perf: refine onnx implementation

[larme]

• multiple outputs support
• more detailed input type casting

What does this PR address?

Fixes #(issue)

Before submitting:

• Does the Pull Request follow Conventional Commits specification naming? Here are GitHub's
  guide on how to create a pull request.
• Does the code follow BentoML's code style, both make format and make lint script have passed (instructions)?
• Did you read through contribution guidelines and follow development guidelines?
• Did your changes require updates to the documentation? Have you updated
  those accordingly? Here are documentation guidelines and tips on writting docs.
• Did you write tests to cover your changes?

[codecov]

Codecov Report

Merging #3166 (65ade2a) into main (22bd3f8) will decrease coverage by 0.17%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##             main    #3166      +/-   ##
==========================================
- Coverage   22.49%   22.31%   -0.18%     
==========================================
  Files          97       98       +1     
  Lines        9226     9300      +74     
  Branches     1550     1563      +13     
==========================================
  Hits         2075     2075              
- Misses       7108     7182      +74     
  Partials       43       43              

Impacted Files	Coverage Δ
src/bentoml/_internal/frameworks/onnx.py	0.00% <0.00%> (ø)
src/bentoml/_internal/utils/onnx.py	0.00% <0.00%> (ø)

[larme]

will add api v1 backward computability and more tests later