Add type hints to top-level public API functions (#8079)
## What
Adds precise type hints to the top-level public API functions in
`deepspeed/__init__.py`, addressing #8074:
- `initialize` — annotates the previously-untyped `args`, `mpu`,
`collate_fn`, `config`, `mesh_param`, `config_params` parameters and
adds the return type.
- `init_inference`, `default_inference_config`, `add_config_arguments`,
`tp_model_init`, `set_optimizer_flags`.
## Why
#8074 asks for type annotations on the public API. This is a scoped
first increment covering the top-level `deepspeed.*` entry points (the
functions users call directly), which keeps the diff small and
reviewable; follow-ups can extend to deeper modules.
## Notes
- Annotations only — no runtime behavior change.
- `pre-commit run --files deepspeed/__init__.py` is green (yapf, flake8,
check-torchdist, check-license, codespell).
- `initialize` returns `Tuple[DeepSpeedEngine, Optional[Optimizer],
Optional[DataLoader], Optional[_LRScheduler]]` — both
`DeepSpeedHybridEngine` and `PipelineEngine` subclass `DeepSpeedEngine`.
- Genuinely dynamic / duck-typed parameters (`args`, `mpu`,
`mesh_param`) are annotated `Any`.
Opened as a draft for early feedback on the scope and the chosen
annotations.
Addresses #8074
---------
Signed-off-by: Arun Sharma <sharm485@umn.edu>
Co-authored-by: Olatunji Ruwase <tunji.ruwase@snowflake.com>
