Declare depth compression params in a per-topic namespace

[Kettenhoax]

This change declares the parameters of compressedDepth in the same namespace as done by compressed since 30d0401

Indirectly fixes #89, because the depth_quantization parameter of the publisher was inadvertently assigned to the wrong config field before.
Presumably the quantization parameter remained 0 or uninitialized, which could result in the bad values.

Edit
I used this adapted gazebo depth example to test:
depth_camera_ros.zip

[bmegli]

Indirectly fixes #89, because the depth_quantization parameter of the publisher was inadvertently assigned to the wrong config field before.

ROS1

In ROS1 noetic it was still using the correct value (dynamic reconfigure provided).

image_transport_plugins/compressed_depth_image_transport/src/compressed_depth_publisher.cpp

Lines 69 to 77 in cca6781

	void CompressedDepthPublisher::configCb(Config& config, uint32_t level)
	{
	config_ = config;
	}
	void CompressedDepthPublisher::publish(const sensor_msgs::Image& message, const PublishFn& publish_fn) const
	{
	sensor_msgs::CompressedImage::Ptr compressed_image =
	encodeCompressedDepthImage(message, config_.format, config_.depth_max, config_.depth_quantization, config_.png_level);

ROS2

Looks like the bug you mention was introduced during large API reworking for ROS2:

• https://github.com/ros-perception/image_transport_plugins/pull/31/files#diff-c6bad074b07e07da26c20873fd26f928592551af962f1894b9b5cbdc15a48fb3

Consequence

Presumably the quantization parameter remained 0 or uninitialized, which could result in the bad values.

Uninitialized. Which may often happen to be 0 by pure luck. The intricacies of C++ initialization that lead to this are funny but irrelevant.

Good catch @Kettenhoax

[ijnek]

This PR was included in #145 and merged. @Kettenhoax Apologies I didn't get around to reviewing this PR earlier, but thank you for your contribution!