Fix Pmetrics CSV import and export

[Siel]

Summary

• Fixes the existing Pmetrics reader and writer to preserve complete Data values that Pmetrics CSV can represent.
• Adds Data::as_bytes and Data::from_bytes; write_pmetrics and read_pmetrics use those same paths.
• Writes real observations and doses only: the first occasion has no reset row, and a later occasion marks its initial dose with EVID=4.
• Writes covariates on real event rows at matching times and returns an error for standalone covariate times.
• Writes missing observations as OUT=-99 and rejects an actual observed value of -99.
• Preserves the existing lowercase-header reader behavior and keeps older Pmetrics files and bincode data readable.
• Retains exact covariate source values without changing the serialized covariate layout.

Validation

• cargo fmt --check
• cargo check --all-targets --all-features
• focused Pmetrics tests: 24 passed
• cargo test -p pharmsol --lib: 395 passed
• git diff --check

[github-actions]

Bencher Report

Project	pharmsol
Branch	feature/pmetrics-csv-v1
Testbed	mhovd-pgx

⚠️ WARNING: Truncated view!

The full continuous benchmarking report exceeds the maximum length allowed on this platform.

🐰 View full continuous benchmarking report in Bencher

[mhovd]

Writes missing observations as OUT=-99 and rejects an actual observed value of -99.
I think I prefer that missing observations are written back to the CSV as actual missing values. What do you think?