docs: describe the file stream contract for storage engines

[UlisesGascon]

Documents in StorageEngine.md what a custom engine must do with file.stream:

• always consume it; call file.stream.resume() to discard the rest of a file, never destroy() (busboy cannot continue after a destroyed part stream and the request hangs; see feat: Allow custom storage engines to prematurely terminate stream without calling cb(err) #1277)
• on an aborted request file.stream emits error and close; use stream.pipeline(file.stream, out, cb) so the output stream is destroyed and cb gets the error (see docs: document StorageEngine abort contract (file.stream error/close semantics) #1459, Suggestion: Expose outStream in _handleFile to allow manual stream cleanup on request abort #1300)
• set file.path before writing so multer can call _removeFile for a partially written file

The template is updated accordingly (it used bare pipe() + finish) and was run through a normal and an aborted upload: no leftover file after abort.

[coveralls]

coverage: 99.727%. remained the same — docs/storage-engine-stream into main