fix: preserve literal data in multiline custom tags

[davidfant]

Problem

Multiline custom-tag preprocessing also processes literalTagContent tags. Its blank-line sandwich makes entity-encoded punctuation following a bare URL enter Markdown/GFM parsing. For example, a literal JSON body containing {"websites":["https://example.com"]} arrives with the closing ] still encoded and is no longer valid JSON.

Fix

Exclude explicitly literal tags from Markdown-container preprocessing. Keep existing literal escaping/restoration and normal Markdown custom-tag processing unchanged.

Validation

Added a renderer regression that fails on current main and passes with this change. 47 tests pass across custom-tag Markdown, preprocessing, and literal-content suites. Downstream reproduction: https://github.com/comp-work/monorepo/pull/1896 (private repository), where URL-bearing suggested tasks disappear while URL-free tasks survive. This source change corresponds to the downstream maintained patch; no sanitizer rules are relaxed.

[vercel]

@davidfant is attempting to deploy a commit to the Vercel Team on Vercel.

A member of the Team first needs to authorize it.