(Yeah I'll automate this at some point)

resolves #4477

This PR adds support for dict log arguments and adds (cursor generated)
tests accordingly.

We're surfacing internal SDK errors to users in
#4410.

Regular tox update.

This includes a fix for a new Bottle version which made one of our tests
get stuck in a neverending `while` loop.

New release of `tracerite` (a transitive dependency in the sanic
workflow) causes
[this](https://github.com/getsentry/sentry-python/actions/runs/15735197921/job/44345942053?pr=4493)
to happen.

Related `tracerite` bug report:
sanic-org/tracerite#20 Once this is fixed, we
can unpin.

The soft exit wasn't properly shutting down the thread if another
profiler started up too quickly. This ensures it is reused if possible
but is properly shutdown if needed.

Specifically, the shutdown allowed the profiler 1 cycle before actually
shutting down. If another profiler is started during this cycle, it's
possible the old profiler never shuts down. Resulting in multiple
profilers running.

Fixes #4489

This reverts commit 3f9acc4.

- tracerite 1.12 contained syntax that did not work on Python 3.x
- tracerite 1.13 was then released, containing a fix
- however, on Python 3.8 newest tracerite seems to be using importlib
features that were only added in 3.9 (see below), so still pinning it to
an older version there

```
Traceback:
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/_pytest/python.py:493: in importtestmodule
    mod = import_path(
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/_pytest/pathlib.py:587: in import_path
    importlib.import_module(module_name)
../../.pyenv/versions/3.8.18/lib/python3.8/importlib/__init__.py:127: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
<frozen importlib._bootstrap>:1014: in _gcd_import
    ???
<frozen importlib._bootstrap>:991: in _find_and_load
    ???
<frozen importlib._bootstrap>:961: in _find_and_load_unlocked
    ???
<frozen importlib._bootstrap>:219: in _call_with_frames_removed
    ???
<frozen importlib._bootstrap>:1014: in _gcd_import
    ???
<frozen importlib._bootstrap>:991: in _find_and_load
    ???
<frozen importlib._bootstrap>:975: in _find_and_load_unlocked
    ???
<frozen importlib._bootstrap>:671: in _load_unlocked
    ???
<frozen importlib._bootstrap_external>:843: in exec_module
    ???
<frozen importlib._bootstrap>:219: in _call_with_frames_removed
    ???
tests/integrations/sanic/__init__.py:3: in <module>
    import sanic
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/sanic/__init__.py:6: in <module>
    from sanic.app import Sanic
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/sanic/app.py:58: in <module>
    from sanic.application.state import ApplicationState, ServerStage
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/sanic/application/state.py:13: in <module>
    from sanic.server.async_server import AsyncioServer
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/sanic/server/__init__.py:5: in <module>
    from sanic.server.runners import serve
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/sanic/server/runners.py:6: in <module>
    from sanic.config import Config
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/sanic/config.py:13: in <module>
    from sanic.errorpages import DEFAULT_FORMAT, check_error_format
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/sanic/errorpages.py:27: in <module>
    from sanic.pages.error import ErrorPage
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/sanic/pages/error.py:3: in <module>
    import tracerite.html
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/tracerite/__init__.py:1: in <module>
    from .html import html_traceback
.tox/py3.8-sanic-v24.6/lib/python3.8/site-packages/tracerite/html.py:5: in <module>
    from importlib.resources import files
E   ImportError: cannot import name 'files' from 'importlib.resources' (/Users/ivana/.pyenv/versions/3.8.18/lib/python3.8/importlib/resources.py)
```

It also added performance and aws related files but I want to keep it
simple for now.
This adds:

* quick reference
* testing guide
* project overview
* core architecture
* integration guide

Resolving #4443 requires some changes to this method, but the current
`args`/`kwargs` business makes the method difficult to reason through.

This PR simplifies the logic by listing out the parameters we need to
access, so we don't need to access them through `args` and `kwargs`.

We also cut down on the amount of branching and the amount of variables
(`new_callbacks` vs `existing_callbacks`).

Behavior does not change in this PR; we fix the #4443 bug in #4485,
which is based on this PR
<!-- Describe your PR here -->

---

Thank you for contributing to `sentry-python`! Please add tests to
validate your changes, and lint your code using `tox -e linters`.

Running the test suite on your PR might require maintainer approval.

Add support for AI agents projects using `openai-agents`
(https://pypi.org/project/openai-agents/)

Docs PR is here: getsentry/sentry-docs#14113

This integration:
- records tracing data of agent invocation, tool execution, ai client
requests to LLMs, and handoffs to other agents.
- captures input and output to/from LLMs if `set_default_pii=True`.
- is mostly compatible to the OpenTelememetry `gen_ai` semantic
conventions. (input and output is not compatible because Sentry does not
have Span events. This information is stored in arrays on the Span
attributes.
- Captures errors that happen during agent execution (like problems
during interaction with the LLM.

This integration does not:
- Capture errors during function tool exection because this is very hard
to patch (see comment in the code)

Example span tree in Sentry.io:
![Screenshot 2025-06-24 at 12 15
17](https://github.com/user-attachments/assets/87199067-434f-4bb9-b563-5c6fc18c56cb)

---------

Co-authored-by: Ivana Kellyer <ivana.kellyer@sentry.io>