docs: expand release_candidate2 documentation

[mattloose]

Summary

Reworks and expands the documentation for release_candidate2 so the MkDocs site is the canonical user/developer documentation and first-time users have a much clearer path through the project.

First-run documentation

• rewrites the Home page around clear entry points
• turns Quickstart into a short first-run sequence rather than a second command reference
• makes Installation self-contained
• moves detailed workflow/CLI behaviour into the command reference
• reorders the MkDocs navigation around the user journey
• substantially simplifies the repository README and points it into MkDocs instead of duplicating the full manual

Analysis and developer documentation

• architecture and workflow/data-flow overview
• analysis pipeline overview
• methylation classifier documentation (Sturgeon, NanoDX, PanNanoDX, random forest)
• MGMT promoter methylation documentation
• CNV documentation
• target coverage and variant-analysis documentation
• fusion and ITD structural-event documentation
• CLI job-type documentation updated to match release_candidate2, including ITD and TUCAN

Repository cleanup

• removes the unused git-lfs conda dependency; the current branch has no LFS-tracked paths
• removes obsolete documentation references to private GitHub/GITHUB_TOKEN requirements

Scope

The application and analysis implementation are unchanged. The only non-documentation codebase change is removal of the unused git-lfs package from robin.yml.

The documentation explicitly distinguishes model/heuristic outputs from standalone clinical interpretation and describes the incremental per-sample processing model used by ROBIN.