add Chunk + Fields API with selective decode

[georgeboot]

Summary

Replaces the old DecompressedChunk + Vec<las::Point> reader path with a column-oriented Chunk wrapping las::PointCloud (from gadomski/las-rs#143) and a Fields bitflag that drives laz::DecompressionSelection for layer-level selective decode. The point is to stop decoding layers nobody reads and stop materializing points we're just going to throw away.

Two-tier reader API:

• Simple — fetch_points / query_points / query_points_to_level return Vec<las::Point>. Unchanged shape for scripts and tests.
• Fast — fetch_chunk / query_chunks / query_chunks_to_level take a Fields mask and return Chunks with zero-copy column access. Field accessors return Option so skipped fields can't silently read as zero.

copc-temporal mirrors the same split with query_chunks / query_chunks_by_time alongside the existing query_points variants.

Depends on the updated jacovdbergh/las-rs point-cloud-api branch (pinned to 833cf7a). Bumps the workspace version to 0.2.0-dev since DecompressedChunk and the old read_points* methods are gone.

Before / after

Fetching only xyz + rgb for a bounding box:

// before — decoded every layer, materialized every point
let points = reader.query_points(&bbox).await?;
for p in &points {
    do_something(p.x, p.y, p.z, p.color);
}

// after — selective decode, zero-copy column walks
use copc_streaming::Fields;

let chunks = reader.query_chunks(&bbox, Fields::Z | Fields::RGB).await?;
for chunk in &chunks {
    let positions = chunk.positions().unwrap();
    let rgb = chunk.rgb().unwrap();
    for (pos, rgb) in positions.zip(rgb) {
        do_something(pos[0], pos[1], pos[2], Some(rgb));
    }
}

Temporal query that intersects a spatial and time window:

// before — materialized every point in the candidate range
let points = temporal.query_points(&mut reader, &bbox, start, end).await?;

// after (fast path) — walk columns, filter, materialize only matches
use copc_streaming::Fields;
use copc_temporal::indices_in_time_range;

let chunks = temporal
    .query_chunks(&mut reader, &bbox, start, end, Fields::Z | Fields::GPS_TIME)
    .await?;

for (chunk, candidate_range) in &chunks {
    let in_time = indices_in_time_range(chunk, start, end).unwrap();
    let in_bounds = chunk.indices_in_bounds(&bbox).unwrap();
    // intersect as you see fit, then walk the columns directly
}

// the simple `query_points` still works unchanged if you just want owned points

Test plan

• cargo build --all
• cargo clippy --all --all-targets -- -D warnings
• cargo test --all — 42 unit tests + 1 doctest pass
• cargo doc --no-deps --all — no warnings, no broken links
• cargo semver-checks check-release -p copc-streaming -p copc-temporal — passes against the 0.1.5 baseline