docs(readme): mark scikit-learn→MLJ as planned, not shipped (G38) (#49)

<!-- SPDX-License-Identifier: CC-BY-SA-4.0 -->
## Summary

**G38** — the README "Supported Patterns" table claimed `scikit-learn
pipelines → MLJ.jl` as a translated pattern, but the codegen implements
**no** sklearn→MLJ conversions. Marked that one row as ⚠ Planned to
match reality.

## Changes

- `README.md`: the `scikit-learn` row now reads ⚠ **Planned** — imports
are *detected* (`sk` alias in `src/codegen/parser.rs`) but no
sklearn→MLJ function conversions are mapped in
`src/codegen/julia_gen.rs`.

## RSR Quality Checklist

### Required

- [x] No banned language patterns
- [x] SPDX headers intact (docs-only change)
- [x] No secrets/credentials

### As Applicable

- [x] Documentation updated for user-facing changes (this IS the doc
correction)

## Testing

Ground-truthed the whole table against `src/codegen/julia_gen.rs`:
- **Implemented** (left as-is): `pandas` (15 maps), `numpy` (17),
`scipy` (optimize→Optim, integrate→QuadGK), `matplotlib.pyplot`→Plots
(plot/scatter/hist/…), R `data.frame` via
`dplyr`(10)/`tidyr`(4)/`stats`(5)/`readr`(2).
- **Overclaimed** (fixed): `scikit-learn`→MLJ — zero maps in the
codegen; only the alias is detected.

So only the sklearn row was inaccurate.

---

**Noted for follow-up (not in this PR):** G37 (unrendered
`{{PROJECT_NAME}}`/`{{project}}`/`{{AUTHOR}}` template placeholders in
Containerfile/guix.scm/.github community docs — mostly comment/header
cruft), G34 (`.idr` never typechecked — needs an `idris2 --build` CI
job), G36 (7 empty `src/` subdirs + empty `verification/` scaffolding vs
`phase-1-complete` STATE), G39 (`cargo install` + SPDX on `.rs`/`.zig` —
licence-touchy, owner-gated like Axiom G17). The README itself is `.md`
where the estate default is `.adoc` (format drift, separate from this
content fix).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---
_Generated by [Claude
Code](https://claude.ai/code/session_01UPFC9YQ7g9gc3VnRox42Q1)_

Co-authored-by: Claude <noreply@anthropic.com>

Author: hyperpolymath

README.md

@@ -91,7 +91,7 @@ Julianiser recognises and translates these common data science patterns:
 | `numpy` arrays | Native Julia arrays | Broadcasting, slicing, linear algebra |
 | `scipy.optimize` | `Optim.jl` / Julia stdlib | Minimisation, root-finding, curve fitting |
 | `matplotlib` / `seaborn` | `Plots.jl` / `Makie.jl` | Static and interactive plotting |
-| `scikit-learn` pipelines | `MLJ.jl` | Train/predict/evaluate pattern |
+| `scikit-learn` pipelines | `MLJ.jl` | ⚠ **Planned** — sklearn imports are *detected* (the `sk` alias is recognised in the parser), but no sklearn→MLJ function conversions are mapped yet in the codegen |
 | R `data.frame` / `tibble` | `DataFrames.jl` | dplyr-style verbs mapped to Julia |
 | R `apply` / `sapply` / `lapply` | Julia broadcasting / `map` | Vectorised equivalents |