[do not merge] feat: Span streaming & new span API

[sentrivana]

Introduce a new start_span() API with a simpler and more intuitive signature to eventually replace the original start_span() and start_transaction() APIs.

Additionally, introduce a new streaming mode (sentry_sdk.init(_experiments={"trace_lifecycle": "stream"})) that will send spans as they finish, rather than by transaction.

import sentry_sdk

sentry_sdk.init(
    _experiments={"trace_lifecycle": "stream"},
)

with sentry_sdk.traces.start_span(name="my_span"):
    ...

The new API MUST be used with the new streaming mode, and the old API MUST be used in the legacy non-streaming (static) mode.

Migration guide: getsentry/sentry-docs#16072

Notes

• The diff is huge mostly because I've optimized for easy removal of legacy code in the next major, deliberately duplicating a lot. I'll of course split it up to reviewable PRs once ready.
  • Chose to go with a new file and a new span class so that we can just remove the old Span and drop the new StreamedSpan in tracing.py as a replacement.
• The batcher for spans is a bit different from the logs and metrics batchers because it needs to batch by trace_id (we can't send spans from different traces in the same envelope).

Release Plan

• There will be prereleases for internal testing.
• We'll release the new API in a minor version as opt-in.
• In the next major, we'll drop the legacy API.

Project

https://linear.app/getsentry/project/span-first-sdk-python-727da28dd037/overview

[github-actions]

Semver Impact of This PR

⚪ None (no version bump detected)

📋 Changelog Preview

This is how your changes will appear in the changelog.
Entries from this PR are highlighted with a left border (blockquote style).

---

Bug Fixes 🐛

Span Streaming

• fix(span-streaming): Always preserialize attributes by sentrivana in #5407
• fix(span-streaming): Warn about thread usage if any batcher is active by sentrivana in #5408

Other

• fix(anthropic): Token reporting by alexander-alderman-webb in #5403
• fix(openai): Token reporting by alexander-alderman-webb in #5406
• fix(openai-agents): Inject propagation headers for HostedMCPTool when streaming by alexander-alderman-webb in #5405
• fix: Adapt to new packaging in toxgen by sentrivana in #5382

Internal Changes 🔧

Fastmcp

• test(fastmcp): Wrap prompt in Message by alexander-alderman-webb in #5411
• test(fastmcp): Remove test_fastmcp_without_request_context() by alexander-alderman-webb in #5412
• test(fastmcp): Use AsyncClient for SSE by alexander-alderman-webb in #5400
• test(fastmcp): Use TestClient for Streamable HTTP by alexander-alderman-webb in #5384
• test(fastmcp): Simulate stdio transport with memory streams by alexander-alderman-webb in #5333

Mcp

• test(mcp): Use AsyncClient for SSE by alexander-alderman-webb in #5396
• test(mcp): Use TestClient for Streamable HTTP by alexander-alderman-webb in #5383
• test(mcp): Remove unused stdio helpers by alexander-alderman-webb in #5409

Other

• ci: 🤖 Update test matrix with new releases (02/02) by github-actions in #5413
• ci: Update tox and pin packaging version for tox by alexander-alderman-webb in #5381

Other

• [do not merge] feat: Span streaming & new span API by sentrivana in #5317

---

_{🤖 This preview updates automatically when you update the PR.}