perf(executor): per-ExecutionCtx ArrayKernels snapshot

[lukekim]

What

Resolves ArrayKernels once at ExecutionCtx construction into a crate-private KernelSnapshot (a single ArcSwap::load_full of the execute-parent kernel map), instead of cloning the session and probing the sharded session-variable DashMap on every array node in the execute_until / single-step execution paths.

Why

• Removes a per-array-node session clone + DashMap shard RwLock probe from the hot execution loop.
• Stops holding the session-variable read guard across kernel invocation (previously a plugin kernel touching the session registry could contend/deadlock on the same shard).

Semantics

The registry is session-scoped and mutable via its public register_* methods. An ExecutionCtx sees a point-in-time snapshot taken at construction; later registrations are picked up by the next context (contexts are created per evaluation, e.g. create_execution_ctx() per filter/projection evaluation). Kernel lookup order is unchanged: registered plugin kernels are tried before static execute_parent kernels, with the same (parent, child) hashing.

KernelSnapshot and ArrayKernels::snapshot() are pub(crate) — no new public API.

Testing

• cargo nextest run -p vortex-array — 2789 passed (includes struct_cast_execute_parent_uses_session_plugin, which registers a plugin kernel and then executes through a fresh ctx, covering the register → snapshot → execute path).
• cargo clippy --all-targets -p vortex-array — no diagnostics in changed files.

Split out of #60 (first of two independent changes).

[Copilot]

Pull request overview

This PR optimizes the vortex-array execution hot path by taking a point-in-time snapshot of session-scoped ArrayKernels execute-parent registrations at ExecutionCtx construction, avoiding repeated session-variable lookups and lock probing during execute_until and single-step execution.

Changes:

• Added ArrayKernels::snapshot() and a crate-private KernelSnapshot to capture the current execute-parent kernel registry via a single ArcSwap::load_full().
• Cached Option<KernelSnapshot> inside ExecutionCtx, and updated execute-parent dispatch to consult the snapshot instead of re-reading from the session per array node.
• Updated internal execute-parent helper signatures to accept Option<&KernelSnapshot>.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
vortex-array/src/optimizer/kernels.rs	Introduces KernelSnapshot and ArrayKernels::snapshot() to capture execute-parent kernels as an owned Arc without copying the map.
vortex-array/src/executor.rs	Caches the snapshot in ExecutionCtx and routes execute-parent plugin lookup through the cached snapshot to avoid repeated session probes/locks.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.