Chore: Drop ImageMagik as a dependency (#14440)

* Mark the thumbnail constants Final

* Comment each pdftoppm option

* Feature/Chore: Replaces all usages of ImageMagik's convert binary with a simple Pillow based helper

This removes ImageMagik as a dependency entirly and all documentation is cleaned up from it.
We no longer need to install it, have a policy for it to allow PDFs and no longer shell
out to it either.

* Fix: Save WebP images losslessly when removing the alpha channel

Pillow's default WebP save is lossy, so a WebP with an alpha channel came
back from remove_alpha_channel with slightly shifted pixel values before
it reached OCR. Save WebP lossless; other formats are unchanged.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>

src/documents/tests/test_utils.py

@@ -1,6 +1,11 @@
+from pathlib import Path
+
+import pytest
 import pytest_mock
+from PIL import Image
 
 from documents.utils import QuerySetStream
+from documents.utils import remove_alpha_channel
 
 
 class TestQuerySetStream:
@@ -33,3 +38,99 @@ class TestQuerySetStream:
         # the thing being verified here.
         mock_queryset.count.assert_called_with()
         mock_queryset.iterator.assert_called_once_with(chunk_size=1000)
+
+
+class TestRemoveAlphaChannel:
+    @pytest.mark.parametrize(
+        ("image_format", "mode", "expected_mode", "color", "expected_pixel"),
+        [
+            pytest.param(
+                "PNG",
+                "RGBA",
+                "RGB",
+                (10, 20, 30, 0),
+                (10, 20, 30),
+                id="png-rgba",
+            ),
+            pytest.param(
+                "TIFF",
+                "RGBA",
+                "RGB",
+                (10, 20, 30, 128),
+                (10, 20, 30),
+                id="tiff-rgba",
+            ),
+            pytest.param("PNG", "LA", "L", (40, 0), 40, id="png-la"),
+        ],
+    )
+    def test_drops_alpha_and_keeps_format(
+        self,
+        tmp_path: Path,
+        image_format: str,
+        mode: str,
+        expected_mode: str,
+        color: tuple[int, ...],
+        expected_pixel: int | tuple[int, ...],
+    ) -> None:
+        """
+        GIVEN:
+            - An image with an alpha channel
+        WHEN:
+            - The alpha channel is removed to a destination without an extension
+        THEN:
+            - The destination has the matching opaque mode, the same size and
+              format, and the color values are kept rather than composited
+        """
+        source = tmp_path / "source"
+        Image.new(mode, (8, 6), color).save(source, format=image_format)
+        dest = tmp_path / "no-alpha"
+
+        result = remove_alpha_channel(source, dest)
+
+        assert result == dest
+        with Image.open(dest) as out:
+            assert out.mode == expected_mode
+            assert out.size == (8, 6)
+            assert out.format == image_format
+            assert out.getpixel((0, 0)) == expected_pixel
+
+    def test_keeps_every_frame_of_a_multi_page_tiff(self, tmp_path: Path) -> None:
+        """
+        GIVEN:
+            - A multi-page TIFF where every page has an alpha channel
+        WHEN:
+            - The alpha channel is removed
+        THEN:
+            - All pages are kept and none has an alpha channel
+        """
+        source = tmp_path / "multi.tiff"
+        first = Image.new("RGBA", (8, 6), (255, 0, 0, 50))
+        rest = [Image.new("RGBA", (8, 6), (0, 255, 0, 50)) for _ in range(2)]
+        first.save(source, save_all=True, append_images=rest)
+        dest = tmp_path / "multi-no-alpha.tiff"
+
+        remove_alpha_channel(source, dest)
+
+        with Image.open(dest) as out:
+            assert out.n_frames == 3
+            for index in range(out.n_frames):
+                out.seek(index)
+                assert out.mode == "RGB"
+
+    def test_keeps_dpi(self, tmp_path: Path) -> None:
+        """
+        GIVEN:
+            - An alpha image with a DPI set
+        WHEN:
+            - The alpha channel is removed
+        THEN:
+            - The DPI is unchanged, since OCR scaling depends on it
+        """
+        source = tmp_path / "source.png"
+        Image.new("RGBA", (8, 6)).save(source, dpi=(300, 300))
+        dest = tmp_path / "dest.png"
+
+        remove_alpha_channel(source, dest)
+
+        with Image.open(dest) as out:
+            assert round(out.info["dpi"][0]) == 300