Co-authored-by: Oleks V <comphead@users.noreply.github.com>

Re-enable datafusion's default features (parquet, sql) and add arrow
dependency with the ffi feature so FFI_ArrowArrayStream, ctx.sql, and
register_parquet compile again.

feat: initial seed of Apache DataFusion Java bindings

## Summary

- Removes the entire `protected_branches` block from `.asf.yaml`.
- Removes `.github/workflows/format.yml`.

## Rationale

GitHub Actions has not yet been enabled on the apache repo, so the
`Format` workflow ends in `startup_failure` on every run and never
reports the `Java (spotless)` / `Rust (cargo fmt)` status checks.
Combined with `protected_branches.main.required_status_checks` requiring
those contexts, PRs are unmergeable.

This change clears both sides of the deadlock: the protection block is
removed so future PRs aren't blocked, and the workflow is removed so
we're not carrying a check definition that can't run. Branch protection
and CI can be reintroduced once Actions is enabled by INFRA and the
project's review/CI practices have stabilized.

apache#11)

## Summary

- Sets `github.protected_branches.main: ~` in `.asf.yaml`.

## Rationale

The previous PR (apache#7) removed the `protected_branches` block entirely,
but ASF INFRA only applies settings that are present in `.asf.yaml` — it
does not clear settings when keys are simply omitted. The required
status checks (`Java (spotless)`, `Rust (cargo fmt)`) therefore remain
configured on `main` and continue to block every PR, because the
workflow that would produce those check contexts is also gone.

Explicitly setting `protected_branches.main` to YAML null (`~`) tells
INFRA to clear the existing branch protection rules on `main`. Branch
protection can be reintroduced later once CI is in place.

…ache#9)

## Summary

- Adds `.github/workflows/build.yml` running on pushes to `main` and on
PRs targeting `main`.
- Sets up Temurin JDK 17 and a stable Rust toolchain, caches Maven
(`~/.m2`) and cargo (`native/target`) artifacts, then runs `make test`.

## Rationale

The project currently has no CI: `format.yml` was removed in apache#7 to clear
a deadlock when GitHub Actions wasn't yet enabled, and apache#11 nulled out
the stale branch protection rules left behind. With Actions now in
place, we need at least one workflow that exercises the build on every
PR so regressions surface before merge.

`make test` is the canonical full build entry point documented in
`README.md` and `CONTRIBUTING.md` — it depends on the `native` target
(so it builds the Rust crate first) and then runs the JVM JUnit suite,
which is exactly what CI needs to cover.

## What's in this PR

- `.github/workflows/build.yml`: single `build` job on `ubuntu-latest`,
JDK 17 (Temurin) with Maven cache via `actions/setup-java`, Rust stable
via `dtolnay/rust-toolchain`, cargo cache via `Swatinem/rust-cache`
scoped to the `native` workspace, then `make test`.

## Not in this PR

- Spotless / Apache RAT / clippy / `cargo fmt` checks. These run at
Maven's `verify` phase or via separate cargo commands and aren't wired
into the Makefile yet; they can be added in a follow-up once we decide
whether to extend the Makefile or invoke them directly from CI.
- Matrix builds across OS or JDK versions — start minimal, expand if
needed.

…Parquet (apache#18)

…ache#13)

Rewrite the README around a feature checklist covering the three query
interfaces (SQL, limited DataFrame, DataFusion-Proto logical plans),
data sources, result materialization, and not-yet items. Move build
prerequisites, build/test commands, and the TPC-H test-data section
from README into CONTRIBUTING.md so the README focuses on using the
library rather than building it.