aImErYbArRaUlT
A single shape whose type `python-pptx` can't identify makes MarkItDown fail the entire PPTX, losing every slide. ### Repro Convert a deck containing an unrecognized shape (docling ships one: [`powerpoint_unrecognized_shape.pptx`](https://github.com/docling-project/docling/blob/main/tests/data/pptx/sources/powerpoint_unrecognized_shape.pptx), slide 2 has a shape named `BadShape`). It raises and returns nothing, though the other two slides are ordinary: ``` PptxConverter threw NotImplementedError with message: Shape instance of unrecognized shape type ``` `PptxConverter` reads `shape.shape_type` (in `_is_picture`, `_is_table`, and the group-shape check); for an unmapped shape that access raises and propagates out of `convert()`. ### Expected One unreadable shape should be skipped and the rest of the deck converted, in the same spirit as the tolerance fixes already merged for `PptxConverter`: #2059 (tolerate `None` shape/notes text), #2194 (chart title without a text frame), and #2408 (top-zero sort). The PR is written and tested; it adds a small `_shape_type_or_none()` helper that returns the type or `None` on `NotImplementedError`, so the shape falls through as "none of the handled types" and its own text is still recovered.