Change in coreboot[master]: nb/intel/sandybridge: Fix programming command timings

[Nico Huber (Code Review)]

Attention is currently required from:  Felix Singer, Angel Pons, Arthur Heymans, Patrick Rudolph.
Nico Huber has posted comments on this change. ( https://review.coreboot.org/c/coreboot/+/49287 )

Change subject: nb/intel/sandybridge: Fix programming command timings
......................................................................

Patch Set 1: Code-Review-1

(5 comments)

Commit Message:

https://review.coreboot.org/c/coreboot/+/49287/comment/324ac70a_1eb834dc 
PS1, Line 11: 
That's basically an empty commit message. If you want to avoid documenting
what the interim code did wrong, how about a revert and then let people
review against the original code?

Patchset:

PS1: 
Looks like this is mixing refactorings with a fix for a
commit that mixed so many refactorings together that it
broke things unnoticed... please stop this mess.

I've also left some `resolved` comments that would make
it much easier to review if they were handled in advance.

File src/northbridge/intel/sandybridge/raminit_common.c:

https://review.coreboot.org/c/coreboot/+/49287/comment/d21c1302_c2c4d89c 
PS1, Line 943: 	FOR_ALL_POPULATED_RANKS {
Please consider making these macros readable before any further reviews.

https://review.coreboot.org/c/coreboot/+/49287/comment/c48d5364_1ed01386 
PS1, Line 939: 	/*
             : 	 * Compute command phase shift as the most negative CCC setting
             : 	 * across all ranks. Use zero if none of the values is negative.
             : 	 */
             : 	FOR_ALL_POPULATED_RANKS {
             : 		cmd_delay = MAX(cmd_delay, -ctrl->timings[channel][slotrank].pi_coding);
             : 	}
             : 
This part looks like a mere refactoring. What's the fix? did I miss something?

https://review.coreboot.org/c/coreboot/+/49287/comment/337a02bc_ad161a8b 
PS1, Line 1000: gdcr_cmd_pi_coding_reg
This type is only used once. If there is no intention to change that, please
please please kill it. The type information is very valuable to the reader
and should only be hidden in a header file as a last resort (if shared by
multiple compilation units). You can also declare it right here as anonymous
union.