[FEATURE] Distinguish markdown heading levels in chat output (sizing/weight), and/or allow customization

Problem

In the desktop/web GUI client, markdown heading levels in assistant responses are nearly visually indistinguishable. H1, H3, and H4 render at almost the same size and weight, and H2 oddly renders in a muted gray, which makes it look less prominent than the levels beneath it. The result is that visual hierarchy is effectively lost and heading structure communicates nothing to the reader.

_Screenshot below: H1/H3/H4 are near-identical in size, and H2 is muted gray (appearing lower in the hierarchy than its own children)._

<img width="1004" height="322" alt="Image" src="https://github.com/user-attachments/assets/9092cb7a-cf21-425e-a3eb-f2673c6519ed" />

Why it matters

Long, structured responses (plans, multi-section explanations, reports) rely on heading hierarchy for scannability. Right now the only reliable contrast levers are bold lead-ins and --- rules, which pushes the assistant away from semantic headings toward visual hacks.

Requests (either would help)

1. Hard-coded variation: ship a clearer default type scale so H1 > H2 > H3 > H4 are visibly distinct in size/weight, and fix H2 so it isn't muted relative to its children.
2. Customization: expose markdown typography (heading sizes/weights, base font) as a user setting, similar to the request in #25234 but for the desktop/web clients.

Environment

Claude Code GUI client (desktop/web). Affects assistant-response markdown rendering specifically.

Related

• #25234 (closed, JetBrains-scoped customizable markdown styling)
• #60230 (content-type color coding for assistant output)
• #10291 (visual separator between responses)
• #26390 (terminal renderer drops GFM features)