Fix srcsetWidth methods to properly handle retina variants

[lukastransom]

Description

The getSrcsetSubsetArray() method fails when retina variants (2x, 3x) are configured, causing srcsetWidth(), srcsetWidthWebP(), and related methods to return wrong variants.

Index-based mapping between variantSourceWidths (duplicated for retina) and sorted optimizedImageUrls keys breaks when looking up variants.

Implemented Fix

Replace fragile index mapping with direct width analysis:

1. Iterate through actual widths in the URL set
2. Detect source width by checking retina multipliers (1x/2x/3x)
3. Filter based on source width comparisons
4. Add DPR parameter to control retina inclusion

What's Been Tested

• ✅ All 6 methods: srcsetWidth*(), srcsetMinWidth*(), srcsetMaxWidth*()
• ✅ Both DPR modes: width descriptors (w) vs density descriptors (x)

Configuration: 5 base widths (360, 442, 466, 559, 573) × 2 retina variants = 10 total

Before Fix (Broken)

{{ optimizedImages.srcsetWidthWebP(559, true) }}
→ .../884px/...webp 1x, .../932px/...webp 1x  {# Wrong variants! #}

After Fix (Correct)

{{ optimizedImages.srcsetWidthWebP(559, true) }}
→ .../559px/...webp 1x, .../1118px/...webp 2x  {# Correct! #}

{{ optimizedImages.srcsetWidthWebP(559, false) }}
→ .../559px/...webp 559w                       {# Only 1x variant #}

{{ optimizedImages.srcsetMaxWidth(466, true) }}
→ 360px@1x, 720px@2x, 442px@1x, 884px@2x, 466px@1x, 932px@2x

Related Issues

• srcsetWidth methods return incorrect variants when retina sizes are configured #431

[khalwat]

I'm so sorry, this fell off of my radar. I'm planning to merge it in and backport it this week.

Are there any changes since your initial PR?

[lukastransom]

This is the same patch I've been using for my 2 latest projects w/o issue. No additional changes on my end!