feat(synthesize): use string-length stats for unstructured text columns

[jqnatividad]

Summary

• synthesize now reads min_length / max_length / avg_length / stddev_length from qsv stats and uses them to shape generated string lengths for unstructured text columns (lorem_*, free_text, no-faker fallback).
• Target length per row is drawn from Normal(avg, stddev) clamped to [min, max], falling back to Uniform(min, max) when stddev_length is absent or zero. Generated text is then truncated (UTF-8-safe, never padded) to the target.
• Structured fakers (email, name, uuid, phone, address parts, …) keep length_stats = None so their semantic format is preserved — truncating an email or UUID would corrupt it.
• Frequency-enumerated and pooled real-value paths are unaffected.
• Box-Muller is used inline to avoid adding a rand_distr dependency.
• Behavior is automatic when length stats are present — no new CLI flag.

End-to-end check on a source column with min=11, max=30, avg=20.75, stddev=7.05 produced synthesized output with min=11, max=30, avg=20.26, stddev=6.01 at n=200.

Test plan

• cargo +nightly fmt
• cargo build --locked --bin qsv -F all_features
• cargo test -F all_features synthesize — 35 unit + 13 integration tests pass (new: 9 unit + 1 integration)
• cargo clippy -F all_features --bin qsv -- -D warnings clean
• qsv --generate-help-md regenerated docs/help/synthesize.md
• Manual smoke test: stats → synthesize → stats round-trip reproduces source length distribution

🤖 Generated with Claude Code

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 43 complexity

Metric	Results
Complexity	43

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

[Copilot]

Pull request overview

This PR teaches qsv synthesize to honor per-column character-length statistics emitted by qsv stats (min_length / max_length / avg_length / stddev_length). For unstructured text columns (lorem_*, free_text, no-faker fallback), generated values are now truncated to a target length sampled from Normal(avg, stddev) clamped to [min, max] (uniform fallback when stddev is missing/zero). Structured fakers (email/uuid/phone/etc.) are intentionally excluded so their semantic format is preserved.

Changes:

• Add LengthStats parsed from StatsRecord::addl_cols and a is_unstructured/UNSTRUCTURED_CONTENT_TYPES gate to decide where length stats apply.
• Wire length_stats into ColumnGenerator::{Faker, LoremFallback} and add sample_target_length (inline Box-Muller) + truncate_to_chars UTF-8-safe truncation invoked in next().
• New unit tests for parsing, sampling, and generator integration (including unstructured pooled vs. structured pooled), plus a new end-to-end integration test for free-text length stats.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
src/cmd/synthesize/generator.rs	Core implementation: LengthStats, sampler, truncation, and integration into the Faker and LoremFallback variants; extensive unit tests.
src/cmd/synthesize/mod.rs	USAGE doc paragraph describing the new length-stats behavior.
docs/help/synthesize.md	Generated help mirroring the USAGE doc paragraph.
tests/test_synthesize.rs	Adds an end-to-end test that stats → synthesize produces blurbs within the source length window.