[PyTorch] Docs: fine-grained quantization recipes (rework of NVIDIA#3336)

[pggPL]

Description

Proposed rework of the docs from NVIDIA#3336 ([PyTorch] Document heterogeneous quantization recipes).

The base branch fine_grained_recipe_docs_base is the PR branch (negvet/fine_grained_recipe_docs @ c3b0398) merged with current main, so the diff here shows only the docs changes on top of NVIDIA#3336.

Rendered preview (NVIDIA network): http://10.64.50.114:5678/gh_fine_grained_recipe_docs/features/low_precision_training/fine_grained_quantization/fine_grained_quantization.html

Type of change

• Documentation change (change only to the documentation, either a fix or a new content)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Infra/Build change
• Code refactoring

Changes

• Rename heterogeneous_quantization -> fine_grained_quantization (feature page + example) and rewrite the feature page in the style of the rest of the Features section: code in tabs/boxes, examples first, less defensive text about the experimental API.
• Add diagrams: precision assignments per module/GEMM, HybridQuantizer data flow, per-GEMM operand mapping, columnwise-source choices. SVGs use the shared diagram-colors.css classes and render correctly in dark mode.
• Add a Quantizer section (PyTorch/JAX tabs) to each recipe page: delayed scaling, current scaling, blockwise scaling, MXFP8, NVFP4; extend the introduction and performance considerations accordingly.
• Simplify the quantizer-factory paragraph and generalize the fallback-path description (per review feedback).
• Shorten the example script; drop the duplicate entry from the top-level docs/index.rst.

Checklist:

• I have read and followed the contributing guidelines
• The functionality is complete
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes