Compat refresh for modern PyTorch/Python + multi-task batch_size=1 fixes

[shenweichen]

Summary

This PR refreshes DeepCTR-Torch compatibility for modern environments and fixes multi-task edge cases observed on newer PyTorch/scikit-learn stacks.

Core compatibility updates

• remove TensorFlow hard dependency from install requirements
• replace TensorFlow private callback imports with native deepctr_torch.callbacks implementations
• update package constraints and metadata for modern Python/PyTorch support
• improve docs/RTD configuration and markdown parser support

Runtime/training fixes

• fix metric input handling for multi-task outputs in BaseModel evaluation/logging path
• fix batch_size=1 multi-task training shape issues by avoiding unsafe global squeeze() in fit path
• keep expert-gating outputs shape-safe in MMOE/PLE (squeeze(1) instead of global squeeze)
• initialize linear accumulator on input tensor device to avoid cross-device mismatch under multi-GPU setups

CI/tests

• add/adjust smoke tests for examples in CI
• add regression coverage for multi-task batch_size=1 behavior

Linked Issues

Closes #281
Closes #288
Closes #303
Closes #305
Closes #263
Closes #309
Refs #301

Validation

• python -m pytest -q tests/models/multitask/MMOE_test.py tests/models/multitask/PLE_test.py -k batch_size_one
• python -m pytest -q tests/models/multitask

[codecov]

Codecov Report

❌ Patch coverage is 98.46154% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.19%. Comparing base (f685425) to head (f7cc1e0).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
deepctr_torch/models/basemodel.py	93.54%	2 Missing ⚠️
deepctr_torch/callbacks.py	99.37%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #310      +/-   ##
==========================================
+ Coverage   92.56%   93.19%   +0.62%     
==========================================
  Files          30       35       +5     
  Lines        2274     2483     +209     
==========================================
+ Hits         2105     2314     +209     
  Misses        169      169              

Flag	Coverage Δ
pytest	93.19% <98.46%> (+0.62%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.