Add context-aware alt text to 36 images across 20 MDX files flagged by mint a11y. Descriptions are derived from each image's caption and surrounding step text. Part of fosrl/docs-v2#116.