## Refactor FRASER pipeline with separated R scripts and improved job management

### Summary

Major refactoring of the FRASER aberrant splicing analysis pipeline, including version bump from 0.2.5 to 1.0.0.

### Changes

#### FRASER Pipeline Refactoring (`src/rdrnaseq/jobs/fraser.py`)
- **Separated R scripts**: Broke monolithic FRASER R code into modular scripts that are baked into and retrieved from a Fraser Image : populationgenomics/images#272
  - `fraser_init.R` - Initialize FRASER object
  - `fraser_count_split.R` - Count split reads per sample
  - `fraser_merge_split.R` - Merge split read counts across samples
  - `fraser_count_non_split.R` - Count non-split reads per sample
  - `fraser_merge_non_split.R` - Merge non-split read counts
  - `fraser_join_counts.R` - Join all count matrices
  - `fraser_analysis.R` - Run FRASER analysis and generate results

- **New `get_fraser_job()` helper function**: Centralized job creation with consistent resource allocation based on `MachineType` (HIGHMEM/STANDARD)

- **Configurable job resources**: Added config-driven storage and CPU allocation:
  - `cohort_job_resources`: For multi-sample jobs (base_storage_gb, per_bam_storage, ncpu)
  - `sample_job_resources`: For single-sample jobs

- **Improved job dependency management**: Fixed job dependency chains to ensure proper execution order across pipeline stages

- **File caching fixes**: Implemented proper output existence checking to skip completed steps

- **Storage optimization**: Intermediate results now saved to temporary GCS storage, with significant results saved permanently

#### Stage Updates (`src/rdrnaseq/stages.py`)
- Updated Fraser stage with new analysis keys: `Rds_data`, `seqr_data`, `sig_results`
- Improved CRAM→BAM conversion tracking to avoid duplicate conversions across stages
- Added `samples_needing_bams` tracking dict for cross-stage job dependencies

#### Configuration (`src/rdrnaseq/config_template.toml`)
- Added `cohort_job_resources` section for cohort-level job configuration
- Added `sample_job_resources` section for sample-level job configuration

#### Cleanup
- Removed deprecated `picard.py` module (465 lines deleted)
- Updated README with current pipeline usage
- Updated Docker image reference
- Version bump: 0.2.5 → 1.0.0

Co-authored-by: Matt Welland <mattwellie@gmail.com>