Bring in more timers for initialization

[ekluzek]

In order to assess scalability issues in initialization we need more timers especially for clm_init2 and in the lnd_set_domain code. Put timers for whole subroutines around the calls, and have internal timers for sections of a subroutine be inclusive of the entire subroutine.

Also remove some of the timers that are around inconsequential parts of the code (time is so low that it can be ignored).

[ekluzek]

I've started doing this in PR #3335 on the decomp init study branch. I'll cherry pick the commits and send them to b4b-dev.

One thing I learned from John Dennis and from doing this work is that the timers should be constructed so the sub timers inside of a subroutine should add up to the timer around the entire call. You sometimes think you can guess which parts are expensive and only have timers around those sections -- but in practice you'll guess wrong or things change and you'll need more timers that show all parts of the code beneath the higher level timer.