Cleaner logging tests

[mxgrey]

This PR takes advantage of rcutils_logging_set_output_handler to override the way log outputs are published within the logging test function. We prevent the messages from being printed and instead collect them in way that the test can access them and validate what is being produced.

This has two big advantages:

• We no longer get false "Error" and "Warning" messages printing out to the terminal when running these tests
• We can explicitly validate that the logger is behaving the way we expect

This branch builds directly off of #7, so that should be merged before this is reviewed.

[mxgrey]

I just realized a major issue with some of the changes in this PR.

When you specify a custom log output handler function to rcutils, your handler function will be given the format string ("%s") along with the variadic arguments that the user passed in to fill into the format string.

The variadic arguments are pretty much indecipherable from the Rust side, so in order to make sense of the log messages on the rclrs side, I skipped passing ("%s", msg) to rcutils and instead simplified it to just passing in msg with no variadic arguments.

This is fine in our tests since we can choose to just look at the message and ignore the variadic arguments. However this will be a problem when we're not testing because it's possible for someone to pass a string with a % character into their log message, and when rcl processes that according to its normal formatting rules, it can result in a segmentation fault or other undefined behavior.

I'll need to tweak the PR to have slightly different behavior when compiled for tests vs compiled for real. A better solution would be to find a way to format a string from Rust given C variadic arguments, but I expect that to be an unreasonably difficult thing to do.

[mxgrey]

The safety issue is fixed by 89809d1, so this PR should be good to go.

[geoff-imdex]

As always, lots of great stuff (definitely helping my Rust learnings!). One extremely pedantic comment for adding in the standard safety stuff.

[mxgrey]

Based on some discussions in the working group, I've updated interval to be called throttle instead, aligning the terminology with the other ROS client libraries.

[mxgrey]

I've also merged in the latest main branch from upstream. I think this PR should be good to go. Hopefully this will be the last one needed for the maintainer review.