[Feature Request] Make Remote OCR respect `PAPERLESS_OCR_MODE` and local preprocessing options

Description Is your feature request related to a problem? Please describe. Currently, when configuring Paperless-ngx to use a remote OCR engine (e.g., PAPERLESS_REMOTE_OCR_ENGINE=azureai), the remote parser unconditionally processes the document and generates an archive file. This behavior completely bypasses the standard OCR control mechanisms that are available for the default Tesseract engine. See also: [ Specifically, this causes two main issues: 1. Duplicate Text Layers: Born-digital PDFs or documents that already have a text layer are sent to the remote OCR engine anyway. Because PAPERLESS_OCR_MODE=auto or skip is ignored by the remote parser, the remote engine generates a secondary, overlapping text layer on top of the existing one. 2. Ignored Preprocessing: Preprocessing options such as PAPERLESS_OCR_CLEAN, PAPERLESS_OCR_DESKEW, and PAPERLESS_OCR_COLOR_CONVERSION_STRATEGY are not applied to the document when routed through the remote OCR parser. While it was …