Remove assumptions and checks around QoS policies that don't affect compatibility

[emersonknapp]

Part of #125
Depends on ros2/rclcpp#1049

Some of the policies in QoS profiles don't affect compatibility, only local behavior, so it's not worth comparing those values. The different RMW implementations report these values inconsistently currently, so though it is useful to record everything, it is not meaningful to test for specific values.

[thomas-moulard]

I think keeping a trace of how things have been recorded is valuable for debugging later. If we have this feature already, I find it a bit counter-productive to throw this data away.

Example use case: "rosbag2 has troubles recording everything, and is dropping messages? oh yeah the history depth was set to 1 erroneously, let's change this behavior". I'm not sure if we already have a rosbag2 info like in ros1, but that's something we could pretty print there.

I understand that some RMW implementations may not be "there yet" in terms of reporting data correctly, but we should continue pushing them to get better, instead of trying to workaround their quirks, unless there is a clear loss for our customers.

I think what's missing maybe from your PR message, is why you're doing this? What problem are you fixing here for users?

[emersonknapp]

@thomas-moulard yeah I think you're right. My main problem had to do with checking for equality. The real solution here is not to check for exact values, but continue recording them as yes, they are potentially-useful information. I've rewritten the PR to reflect this new approach.

[emersonknapp]

• Linux
• Linux-aarch64
• macOS
• Windows

[emersonknapp]

I am closing this PR in favor of a more complete approach that will cause better behavior on average (with a more accurate branch name)