feat(multiplex-quant): expose --sample-bc-ori as a CLI override

[an-altosian]

Summary

Adds --sample-bc-ori {forward,reverse} to simpleaf multiplex-quant, completing the precedence pattern that --expected-ori and --geometry already follow:

1. user-supplied CLI value (this PR)
2. chemistry preset's declared sample_bc_ori
3. omit the flag (alevin-fry default = forward)

Resolves the use case in #198: cycle-plan variants like 10x Flex Configuration B (R1=28 / R2=90) where the sample BC is read from the opposite strand. Users can now run such variants with --chemistry 10x-flexv2-gex-3p --geometry '...' --sample-bc-ori forward without adding a new preset entry.

Implementation notes

• CLI vocabulary: forward / reverse — matches the chemistry preset JSON ("sample_bc_ori": "forward" | "reverse") and alevin-fry's own --sample-bc-ori flag (verified in alevin-fry main.rs).
• No translation layer: The CLI value is forwarded verbatim to alevin-fry's generate-permit-list --sample-bc-ori.
• Preset JSON untouched: When the override comes from the chemistry preset's Option<String> field, it passes through verbatim. No breaking change to existing chemistries.json.
• No both: alevin-fry's --sample-bc-ori only accepts forward/reverse. No current preset declares both either. Adding both would mean inventing semantics that don't exist downstream.

Validation

End-to-end verified by running the same internal Config B Flex library two ways and confirming byte-identical results:

• Path A (preset-driven): --chemistry 10x-flexv2-gex-3p-config-b (from the companion preset PR feat(chemistries): add 10x-flexv2-gex-3p-config-b preset #201, which declares sample_bc_ori: "forward").
• Path B (CLI-override): --chemistry 10x-flexv2-gex-3p (which declares sample_bc_ori: "reverse" in its JSON) + --sample-bc-ori forward + matching --geometry override at the CLI.

Both runs produced identical per-sample-well cell counts and read counts, identical wall-clock time, and sample_bc_ori: Forward in the resulting sample_info.json. That demonstrates:

1. The CLI value reaches alevin-fry generate-permit-list correctly (the demultiplexing wells light up, instead of producing the noise-floor result the preset's declared reverse would yield on Config B reads).
2. The precedence rule is correct: when CLI and preset disagree, the CLI value wins.

CI status note

The check_formatting job fails on this PR with a SIGILL from alevin-fry --version during test_simpleaf.sh. This is a pre-existing failure on upstream main since commit 52ccc73 (the 0.25.0 release): the bioconda alevin-fry 0.15.0 binary appears to use CPU instructions not supported by the GHA runner. My PR does not touch any code path involved in the failure (set-paths runs before any --sample-bc-ori code is exercised). Suggested fixes (cargo install from source in CI, or a bioconda rebuild) are out of scope for this PR; happy to send them separately if desired.

Test plan

• cargo check passes locally on the patched source.
• cargo build --release produces a working binary that runs through multiplex-quant.
• End-to-end run with --sample-bc-ori forward on a real Config B Flex library yields realistic per-sample-well cell counts (matches cellranger multi ground truth at the well level).
• Result is byte-identical to the companion preset path in PR feat(chemistries): add 10x-flexv2-gex-3p-config-b preset #201 (rules out silent path divergence).
• sample_bc_ori is recorded in sample_info.json as the value passed at the CLI, confirming it flowed through to alevin-fry generate-permit-list.
• check_formatting CI job (pre-existing SIGILL on upstream main; see CI status note above).

Refs

• Add chemistry preset for GEM-X Flex v2 Configuration B (R1=28 / R2=90) #198 (Config B preset request — this PR is the general fix the second comment proposed)
• feat(chemistries): add 10x-flexv2-gex-3p-config-b preset #201 (companion: adds 10x-flexv2-gex-3p-config-b chemistry preset that exercises the same sample_bc_ori field via the preset path)

[rob-p]

The CI failure is strange. It suggests an illegal instruction on the CI runner (which is pulling alevin-fry from bioconda). I'm not sure what to do about that.