Allow BAM as input for long-reads submission #98

Description of feature

1. Long-read raw read submission (--mode reads)

Samplesheet schema (assets/schema_input_reads.json)

• fastq_1 only accepts FASTQ. Add a new bam column, and require either fastq_1 or bam. A separate column is clearer than widening the fastq_1 pattern, since ENA treats BAM as its own manifest field.
• Only allow bam when platform is OXFORD_NANOPORE or PACBIO_SMRT, and require fastq_2 to be empty for those platforms.
• To confirm: leave out .tar.gz until ENA confirms whether it will still be supported.
  Workflow (workflows/readsubmit.nf)
• Update the samplesheet mapping for the new bam column and record the file type (FASTQ or BAM) in meta.
  Manifest creation (modules/local/create_reads_manifest/main.nf)
• Add BAM support so the manifest gets a BAM line instead of FASTQ lines.
• To confirm: whether create_reads_manifest in mgnify-pipelines-toolkit (1.5.1) supports BAM.

2. Make BAM uploads retry cleanly

• ENA_WEBIN_CLI_WRAPPER only retries on a few exit codes, once. Network failures during upload usually aren't retried. Add a withName block in conf/modules.config that retries on any failure, at least 3 times, with more time per attempt.
• If an upload succeeds but the task fails afterwards, the retry will be rejected by ENA as already submitted. Check that the wrapper treats this as success and recovers the accession; if not, fix upstream.
• The wrapper copies input files into the work directory. Switch to symlinks if Webin-CLI allows it, to save disk and time.
• Optional: expose Webin-CLI's Aspera option for faster uploads (needs Aspera in the container).

3. Raw read tests

• Add a small ONT FASTQ and a small long-read BAM to nf-core/test-datasets.
• Add tests test_reads_long_fastq and test_reads_long_bam.