Put the RAW decoder behind a Decoder trait

FR-RAW-2 says a second decoder may be added for broader camera coverage
without changing callers, and D2 names LibRaw as that second decoder.
Nothing tested the claim: dr_decode was one decoder reached through free
functions, so adding another would have meant editing every caller at
the moment there was most pressure not to.

Decoder is an object-safe trait over bytes: header_bytes, metadata,
orientation, locate_preview, preview and decode. Rawler implements it by
delegating to the existing free functions, so behaviour is unchanged,
and dr_decode::default() hands it out as a &'static dyn Decoder, which
is what the places that start work will name. JPEG recognition, decoding
and completeness checks stay free functions: they are not a RAW
decoder's to vary.

Nothing in the trait takes a path or a SourceRef; the decoder states how
much of a file it needs and where its preview sits, and the caller's
storage fetches that.

core/dr-decode/src/lib.rs

@@ -11,14 +11,20 @@
 //!
 //! Fusing them would force a full decode where a header read suffices, which
 //! is exactly why Lightroom stalls ~2 s per image during culling.
+//!
+//! Callers reach these through the [`Decoder`] trait rather than by name, so a
+//! second decoder can be put behind them without changing any of them
+//! (FR-RAW-2). [`Rawler`] is the one that ships; [`default`] hands it out.
 
 pub mod base_curve;
+mod decoder;
 mod error;
 mod locate;
 mod preview;
 pub mod profile;
 
 pub use base_curve::BaseCurve;
+pub use decoder::{default, Decoder, Rawler};
 pub use error::DecodeError;
 pub use locate::{
     defects, is_complete_jpeg, jpeg_metadata, locate_preview, tiff_metadata, BadLine, BadPixel,