fix: default SamplesPerPixel to 1 when tag is absent

[TomNicholas]

Fixes #318.

Problem

Parsing a TIFF that omits the SamplesPerPixel tag (277) panicked with samples_per_pixel not found, rather than applying the TIFF 6.0 spec default of 1. Single-channel grayscale TIFFs commonly omit the tag — e.g. the PerkinElmer/Phenix microscopy images in the Cell Painting Gallery. Via virtual_tiff/pyo3-async-runtimes the .expect() surfaces as an uncatchable Rust panic on the tokio worker.

Change

src/ifd.rs: default SamplesPerPixel to 1 when the tag is absent, per the spec:

let samples_per_pixel = samples_per_pixel.unwrap_or(1);

The PlanarConfiguration logic immediately below already handles samples_per_pixel == 1, so this is self-contained.

Test

Adds fixtures/other/grayscale_no_samplesperpixel.tif — a hand-built 114-byte 2x2 uint8 grayscale TIFF that omits tag 277 — and a regression test asserting it parses with samples_per_pixel() == 1. The test panics with samples_per_pixel not found without the fix and passes with it.

[kylebarron]

Could we:

1. Put this image fixture in https://github.com/developmentseed/geotiff-test-data instead of here
2. Make the image creation reproducible. You say you hand-edited the TIFF header, and it would be nice to have a record on how that was done. You can see in the existing rasterio_generated directory how there's a Python script that creates the file.

[kylebarron]

This is verbose; just say

Suggested change

	// SamplesPerPixel (tag 277) defaults to 1 when the tag is absent.
	// TIFF 6.0 spec, "SamplesPerPixel ... Default = 1": https://download.osgeo.org/libtiff/doc/TIFF6.pdf
	// Many single-channel grayscale TIFFs (e.g. PerkinElmer/Phenix microscopy) omit it.
	// Samples per pixel defaults to 1
	// https://web.archive.org/web/20240329145232/https://www.awaresystems.be/imaging/tiff/tifftags/samplesperpixel.html

[TomNicholas]

Sorry yes I can make those changes.

For context @ianhi and others are very interested in using VirtualiZarr on a bunch of bioscience data, lots of which is in weird TIFF variants. So this is not GeoTIFF data, and expect to see lots more edges case like this found soon. (I assume that is still within scope for this project?)