Refactor GeoTIFF Phase 5f: extract _encode.py from _writer.py (#2263)

* Refactor GeoTIFF Phase 5f: extract _encode.py from _writer.py (#2260)

Move the strip / tile encode helpers, the photometric resolution and
MinIsWhite inversion helpers, the predictor normalisation, the
``_compression_tag`` mapping, and the streaming ``_compress_block``
helper out of ``_writer.py`` into a new ``_encode.py``. Mirrors the
``_decode.py`` extraction on the read side (PR-G).

``_writer.py`` re-exports every moved name so the existing public
import path (``xrspatial.geotiff._writer.<name>``) and downstream
backends (``_writers/eager.py``, ``_writers/gpu.py``, ``_writers/vrt.py``,
``_write_layout.py``, the test suite) keep working without churn.

``_write_tiled`` and ``_write_stripped`` look up ``_prepare_tile`` and
``_prepare_strip`` through ``_writer`` so the monkeypatch contract pinned
by ``test_gil_friendly_kwarg_1830`` survives the move.

``_writer.py`` drops from 1800 to 1269 lines.

Part of #2211.

* Address review: clarify _encode lazy-lookup docstring and test refs (#2260)

Tighten the module docstring to point at the actual call sites where
the lazy _writer import lives (_write_stripped /
_write_tiled), and cross-reference the test that pins the
monkeypatch contract from the inline comments next to those lookups.

Pure documentation; no behaviour change.

Author: brendancol