fix(trace-viewer): pretty-print JSON without losing precision

[Yury Semikhatsky (yury-s)]

Summary

• Pretty print in the network pane round-tripped bodies through JSON.parse/JSON.stringify, rounding integers beyond Number.MAX_SAFE_INTEGER (e.g. 64-bit ids) and normalizing literals like 1.0 and 1e3.
• Preserve each number's source text via JSON.rawJSON, falling back to the previous behavior on engines without it (Baseline since March 2025: Chrome 114, Firefox 135, Safari 18.4).

References #42261

[Pavel Feldman (pavelfeldman)]

I don't think this is a good mitigation:

• Either we suggest that users turn off application/json formatter for payloads that aren't parsed by the browser, or
• Bring in the codemirror's formatter mod

[github-actions]

Test results for "MCP"

1 failed
❌ [chrome] › mcp/config.spec.ts:203 › chromiumSandbox › --sandbox enables the sandbox @mcp-ubuntu-latest-chrome

8109 passed, 1341 skipped

---

Merge workflow run.

[github-actions]

Hi, I'm the Playwright bot and I took a first look at the CI failures.

🟢 The one failure is a pre-existing flake — this PR looks clear

[chrome] › mcp/config.spec.ts:203 › chromiumSandbox › --sandbox enables the sandbox is the only failure, and it isn't yours. This PR only touches trace-viewer JSON number formatting; the MCP Chromium sandbox launch path is untouched, and across CI the same test flips verdict — it failed on unrelated PR #42294 (sha 4aa9163c) and passed on this branch. Nothing to fix here.

Details

Overall: the merged report lists 8109 passed, 1341 skipped, and a single failure. That failure has no plausible connection to the diff, and it fails on other people's SHAs too — so this PR is green.

Pre-existing flake / infra

• [chrome] › mcp/config.spec.ts:203 › chromiumSandbox › --sandbox enables the sandbox @mcp-ubuntu-latest-chrome — flake. In the aggregated CI results this test flips final verdict across runs: it failed on PR fix(chromium): stop disabling BoundaryEventDispatchTracksNodeRemoval #42294 (sha 4aa9163c, 2026-08-18 17:42) and passed on this PR's own run (sha 4000f92b), same test on a SHA this PR can't be responsible for. The diff here is confined to networkResourceDetails.tsx (pretty-printing JSON via JSON.rawJSON) and its trace-viewer test — it never exercises MCP sandbox launching.

_{Triaged by the Playwright bot - agent run}