treewide: Use deprecations in code rather than comments

[chrysn]

Contribution description

This executes #18561.

It already has the macros and Doxygen stuff, it lacks

• change of all @deprecated to DEPRECATED
• CI time rejection of non-code deprecation (unsure if possible, maybe we still need it for module level deprecations)
• Removal of deprecated functions in the code base, or annotation with -Wno-error=deprecated-declarations if we still want to test the deprecated functions.

Testing procedure

Right now:

• See how CI starts breaking because we -Werror and the one function that got changed is still used in tests.
• Look at Doxygen output on the changed function.

Long term:

• Try using a deprecated function, see things fail.

[chrysn]

Suggested change

	* `__((attribute(deprecated))`, or is ignored.
	* `__attribute_((deprecated))`, or is ignored.

[maribu]

should be even __attribute__, shouldn't it?

[maribu]

If you throw in an -Wno-error=deprecated-declarations, the warning should not be escalated to an error. We may however want to still escalate it to an error with RIOT_CI_BUILD=1 to avoid in-tree users sneaking back in.

[chrysn]

On the severity of errors, I think a good path would be to have Makefiles of ALLOW_DEPRECATED ?= 1, make Murdock set ALLOW_DEPRECATED = 0, and have some tests (those that explicitly include deprecated functions because we don't want these to accidentally break) set ALLOW_DEPRECATED = 1.

---

On the CI side, Murdock went well because it duly rejected the tests that used the deprecated function (even though, as above, these should later be exempt).

Doxygen did not fail but just failed to upload due to timeouts, retrying.

[chrysn]

Sadly, doxygen output breaks completely with the current doxygen version -- deprecation text gets moved into functions list, but the function itself is not present any more at all, and the next function is somehow also missing ... putting this in as "waiting for CI update" because the next Ubuntu hopefully has a better Doxygen. (Things looked good with the 1.9.4 I have here).

[chrysn]

There was some screw-up with branches -- see #18650 for a follow-up PR.