feat(ether): add network footprint support

[lolito93]

Adds the stl to represent a network footprint report and internal support needed to build it. The report describes the buses and their multicast addresses, multicast exclusions, conflicts and the ports used.

A new reporter can now be installed through PreloadApi and invoked by the kernel with a list of bus addresses.

Resolves SEN-1717

[EnriqueParodi]

Hi Manuel, can you please relay my feedback to the author?

The way this is wired is right. The kernel offers a hook and ether installs the reporter, so the kernel does not need to know the component exists.

1. getNetworkFootprint calls the reporter without checking it is set.

   return networkReport_(busAddresses);
   

   The only installer is components/ether/src/component.cpp:166, and mode=barebones builds no components at all (conanfile.py:22). So in a barebones build the reporter is never installed and this calls an empty callable. move_only_function_impl.h:151 guards it with assert, which is compiled out with NDEBUG, so a debug build aborts with a clear message and a release build calls a null pointer.

   Nothing calls getNetworkFootprint yet, so this cannot happen today. That makes it a good moment to guard it, because once there is a caller the problem only shows up in release builds without ether.

2. Installing a second reporter silently replaces the first. Only ether installs one today, so nothing is wrong, but the header does not say the slot holds one.

3. The kernel side has no caller in this PR. I assume a later one uses it. Worth confirming, otherwise the only thing exercising it is the test.

Thanks!

[andressaumell]

Thanks for the review!

1. This guard already exists in the following PR feat(ether): offline network footprint report #143 .
   However, since getNetworkFootprint() itself is introduced in this PR, I agree the guard belongs here.
2. Okay, we can clarify in component_api.h:240 that the kernel stores a single reporter and that installing another one replaces the existing reporter.
3. Yes, this PR only introduces the reporter and its installation hook. It is consumed in feat(ether): offline network footprint report #143 and a following PR for runtime diagnostic.

[andressaumell]

Hi Manuel, can you please relay my feedback to the author?

The way this is wired is right. The kernel offers a hook and ether installs the reporter, so the kernel does not need to know the component exists.

1. getNetworkFootprint calls the reporter without checking it is set.

   return networkReport_(busAddresses);
   

   The only installer is components/ether/src/component.cpp:166, and mode=barebones builds no components at all (conanfile.py:22). So in a barebones build the reporter is never installed and this calls an empty callable. move_only_function_impl.h:151 guards it with assert, which is compiled out with NDEBUG, so a debug build aborts with a clear message and a release build calls a null pointer.
   Nothing calls getNetworkFootprint yet, so this cannot happen today. That makes it a good moment to guard it, because once there is a caller the problem only shows up in release builds without ether.

2. Installing a second reporter silently replaces the first. Only ether installs one today, so nothing is wrong, but the header does not say the slot holds one.

3. The kernel side has no caller in this PR. I assume a later one uses it. Worth confirming, otherwise the only thing exercising it is the test.

Thanks!

@luisgutierrezpereda and I are preparing the changes:

1. Added a guard to getNetworkFootprint() to throw an error when no reporter is installed.
2. Documented that the kernel stores a single reporter and installing another replaces the existing one.