🤖 This PR was created by Repo Assist, continuing work on the Deedle.MicrosoftML integration epic (#676).

Summary

Adds vector column support to Deedle.MicrosoftML — the key missing piece that enables real ML.NET feature-engineering pipelines (e.g. Concatenate, FastTree output columns).

What's included

src/Deedle.MicrosoftML/DataView.fs

The VBuffer(float) round-trip handles the common case where scalar float inputs are fed through transforms like Concatenate — ML.NET keeps the Double type in the output vector.

tests/Deedle.MicrosoftML.Tests/

5 new tests (22 total):

• Schema check: float32 array column → VectorDataViewType(Single, 3)
• Cursor values: VBuffer(float32) getter reads correct values
• Round-trip: float32 array column survives toDataView → ofDataView
• Mixed frame: scalar + vector columns both survive round-trip
• End-to-end: Concatenate transform output readable as float array

Usage

open Deedle
open Deedle.MicrosoftML
open Microsoft.ML

let mlc = MLContext(seed = 1)

// Pass pre-computed feature vectors to ML.NET
let frameWithVectors : Frame(int, string) =
    Frame.ofColumns [
        "Label",    Series.ofValues [ 1.0; 0.0; 1.0 ] :> ISeries(int)
        "Features", Series.ofValues [
            [| 1.0f; 0.2f; 0.8f |]
            [| 0.1f; 0.9f; 0.3f |]
            [| 0.7f; 0.4f; 0.6f |]
        ] :> ISeries(int)
    ]

let dv = Frame.toDataView frameWithVectors
// → Features column has type VectorDataViewType(Single, 3) ✓

// Use Concatenate to build a feature column, then get result back as Deedle Frame
let pipe   = mlc.Transforms.Concatenate("Features", [| "A"; "B"; "C" |])
let result = Pipeline.fitTransform pipe myScalarFrame
let feats  = result.GetColumn(float array)("Features")
```

## Test status

```
✅ 22 / 22 tests pass (Deedle.MicrosoftML.Tests)

Closes part of #676 (vector column support from the Future Work list)

• Fixes [Repo Assist] Epic: Deedle.MicrosoftML integration package #676

Generated by Repo Assist for issue #676 · ◷

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@30f2254f2a7a944da1224df45d181a3f8faefd0d