Always evaluate BT_ASSERT(); add BT_ASSERT_DBG() for debug mode only

Always evaluate BT_ASSERT(); add BT_ASSERT_DBG() for debug mode only

This patch makes BT_ASSERT() always evaluate its condition.

The goal is to make a lot of _slow path_ assertions evaluated whatever
the debug or non-debug mode. This will help detect project programming
errors and hopefully end users will report them.

What used to be BT_ASSERT() is now BT_ASSERT_DBG(), that is, only
evaluated in debug mode. This is similar to how BT_ASSERT_PRE() is
always evaluated while BT_ASSERT_PRE_DEV() is only evaluated in
developer mode.

I went over each single BT_ASSERT() statement in `src` and decided
whether to use BT_ASSERT() or BT_ASSERT_DBG(). My strategy is similar
to what we do for BT_ASSERT_PRE(), that is:

* Use BT_ASSERT_DBG() in anything potentially executed once or more per
  _event_ message.

  Other messages occur so infrequently compared to event messages that
  we don't care.

* Use BT_ASSERT_DBG() in property getters and object borrowing
  functions.

  We don't know how frequently the user can call those, so we don't take
  any chance.

Everything else uses BT_ASSERT(), for example:

* In the library and plugins, everything related to metadata objects.
* Everything related to graph topology.
* Everything in the CLI.
* Network communication functions in `src.ctf.lttng-live`.

I believe some BT_ASSERT_DBG() statements could still be converted
BT_ASSERT(), but this patch is a good starting point.

I left the whole CTF writer code with BT_ASSERT_DBG() for the moment
as this library is not a priority.

All the tests use BT_ASSERT() because they are not an end user use case;
we don't care if they are less efficient in production mode.

This change does not seem to affect the production build's performance;
I compared, before and after, and I do not get a run time difference
that's greater than the observed measurement error.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Ia74951a39b1fcc79579661562f6a98ed208fd9bb
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2217
Tested-by: jenkins <jenkins@lttng.org>