Fix error reporting formatting (#98)

The stacktrace for an error was formatted as:

    ${message}
    ${stacktrace}

I'm pretty sure I saw that documented in Google Cloud's documentation
when I initially implemented this feature, but I can't find it anymore
and the current documentation from
https://cloud.google.com/error-reporting/docs/formatting-error-messages#log-error
now says:

> To log an error event that is a stack trace, write the error event as
> one of these types:
> [...]
>
> * A `jsonPayload` that includes a `message`, `stack_trace`, or
>   `exception` field.
>
>   You can specify more than one of those fields. If more than one of
>   those fields is specified, then the order of evaluation is: `stack_trace`,
>   then `exception`, and then `message`.
>
>   If the `message` field is evaluated and if it isn't empty, then the
>   stack trace is captured only when the field contains a stack trace in
>   one of the supported programming language formats. The stack trace isn't
>   captured by Error Reporting when an unsupported format is used.
>
>   If your error event is formatted as a `ReportedErrorEvent` object,
>   then copy its fields to the `jsonPayload`. For more information and an
>   example, see Log an error that is formatted as a ReportedErrorEvent
>   object.

There's no mention of this weird `${message}\n${stacktrace}` anymore, so
let's get rid of it.

Allow to override the log level while logging exceptions (#79)

While logging information during an exception `except` block, instead of
overwriting the log level with `CRITICAL`, the log level used by the
caller will be set instead.

This allows the caller to use:

```python
try:
    raise ValueError()
except Exception as exc:
    logger.info(f"I was expecting this error: {exc}")
```

The log message will be logged as `INFO` level, whereas it was logged as
`CRITICAL` before.

If an `exception` attribute is still passed to the logger, as in
`logger.info("something", exception=exc)`, the log level will be `INFO`
as expected but the error will also still be correctly reported (in
Error Reporting + the bubble in Cloud Logging)

Document `structlog_gcp.build_processors` (#44)

Build/test with multiple Python versions (#40)

Use uv to install dev dependencies and build (#38)

Split out the JSON renderer from the Cloud Logging formatter (#30)

This separates the concern a bit more and allow to configure specific
JSON renderer when there's a need to.

Fix fetching the version during release

Actually deploy on releases

Release v0.2.1

Release v0.2.0