fix(client): allow large sse events

[harshmathurx]

Closes #36343

Issue for this PR

Closes #36343

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

Raises the generated Promise client's SSE frame buffer from 1 MiB to 32 MiB so supported large tool success events, such as image read events, do not disconnect /api/event subscribers with MalformedResponse.

The new ceiling still bounds malformed/unbounded SSE frames, but aligns the client with V2's supported media ingest path where a valid encoded image event can exceed 1 MiB.

How did you verify your code works?

• Added a Promise client regression test that streams a >1 MiB SSE event through event.subscribe.
• Ran bun test test/promise.test.ts from packages/client.
• Ran git diff --check.
• Attempted bun run generate from packages/client; blocked locally by dependency resolution: Cannot find module '@effect/platform/Cookies' from @effect/platform-node.
• Attempted bun test test/generate.test.ts from packages/httpapi-codegen; blocked locally because this worktree has no installed effect package.

Screenshots / recordings

N/A

Checklist

• I have tested my changes locally where possible
• I have not included unrelated changes in this PR

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

I found one potentially related PR:

PR #33289: fix(app): prevent web client freeze from delta event bursts and SSE reconnect loops
#33289

This PR addresses SSE-related issues with event handling and reconnect loops, which could be tangentially related to the buffer size issue being fixed in PR #36368. However, it appears to be focused on a different aspect (preventing freezes and reconnect loops) rather than directly addressing the buffer size limitation for large SSE events.

The current PR (#36368) is specifically about raising the SSE frame buffer from 1 MiB to 32 MiB to support large tool success events. This appears to be a targeted fix distinct from the reconnect loop handling in #33289.

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍

[harshmathurx]

Hi @kitlangton — pinging for a review on this PR. It references #36343 and the automated checks are passing, but the "needs:issue" label appears stuck because the GitHub-linked issue is not showing up in the sidebar. Could you confirm the issue is linked (or manually link it) and take a look when you have a moment? Thanks!