Misinterpretation of some BMP files (A and B channels swapped?)

### Problem

OIIO misinterprets https://raw.githubusercontent.com/NatronGitHub/Natron-Tests/master/TestImageBMP/input.bmp

**Expected behavior:**

R=G=B and A=1 everywhere

This is what ImageMagick, the macOS finder, and Xee application see.

**Actual behavior:**
The blue and alpha channels seem to be swapped.

### Steps to Reproduce

(`convert` is from ImageMagick 7)
```
iconvert input.bmp output-oiio.tif
convert input.bmp output-imagemagick.tif
```

### Versions

* OIIO branch/version: 1.8.9
* OS: macOS
* C++ compiler: clang

Reproduce the report with input.bmp using `iconvert input.bmp output-oiio.tif` and compare it with the ImageMagick conversion. Trace the BMP decoding path and verify the channel values against the expected R=G=B and A=1 result; the completed work should preserve the correct channel order and include a regression check for this image.

Written by the indexing model from the issue text.