fix EXIF-rotated images shown cropped in previews, closes #4482

JPEGs with EXIF orientation 5/6/7/8 were rendered cropped in the
mediamanager detail view, the image diff view and the fullscreen
deta

fix EXIF-rotated images shown cropped in previews, closes #4482

JPEGs with EXIF orientation 5/6/7/8 were rendered cropped in the
mediamanager detail view, the image diff view and the fullscreen
detail page: getimagesize() / JpegMeta report raw (rotation-unaware)
pixel dimensions, and passing both w and h to fetch.php defaults to
center-crop.

- Bump splitbrain/slika to 1.1, which ships ImageInfo: a rotation-
aware, metadata-only dimension simulator mirroring Adapter's
fluent API (autorotate/resize/crop).
- Add fit=1 to fetch.php: when both w and h are given, route to
media_resize_image() (bbox fit) instead of media_crop_image().
Token hashes only (id, w, h) so existing tokens stay valid.
- Add MediaFile::getDisplayDimensions($w, $h, $crop) which delegates
to ImageInfo and returns the dims fetch.php would produce.
- Add Display::getDetailHtml() and retire media_preview() / the old
media_image_preview_size() helper. media_tab_view and MediaDiff
now share the Display-based renderer.
- Rewrite tpl_img() (lib/tpl/*/detail.php) to use MediaFile dims +
fit=1 URL; drops the manual ratio math.
- Tests: MediaFileTest covers the dims math, DisplayTest covers the
detail-view HTML (rotated dims, rev-vs-timestamp URL selection,
structure), fetch_imagetoken gains a fit-token compat test.
Fixture _test/data/media/wiki/exif-orient-6.jpg: 20x30 JPEG with
EXIF orientation 6.

show more ...

strip any trailing dots when resolving IDs

fixes #4072

Move PSR compliant tests to the apropriate namespaces

Having "inc" in the namespace is awkward. Instead the test class
namespaces now correspond to their real class namespaces.

With further refacto

Move PSR compliant tests to the apropriate namespaces

Having "inc" in the namespace is awkward. Instead the test class
namespaces now correspond to their real class namespaces.

With further refactorings we should get rid of most of the stuff in
tests/inc

this is a continuation of #3812

show more ...