Hist fields REPRODUCTIVE1N_TO_FOOD_PERHARV and _ANN lose their suffixes in LREPR* tests

[slevis-lmwg]

Brief summary of bug

Two fincl3 history fields from the /crop testmods,
REPRODUCTIVE1N_TO_FOOD_PERHARV and REPRODUCTIVE1N_TO_FOOD_ANN,
lose their suffixes in LREPR* tests and become REPRODUCTIVE1N_TO_FOOD.

This occurs in cime_config/SystemTests/lreprstruct.py
in line user_nl_clm_text = user_nl_clm_text.replace(. The glitch occurs (for an undetermined reason as of this issue's opening) when this line attempts to replace GRAIN with REPRODUCTIVE1 and REPRODUCTIVE2, and REPRODUCTIVE1 ends up losing the suffixes _PERHARV and _ANN.

The problem surfaced in PR #2445 that separates instantaneous from non-inst. fields because:

1. the former fields (with suffixes) are instantaneous, while the latter is "time: mean" and so
2. the LREPR* tests generated h2a files for base and not for grain1 causing a base_grain1 comparison failure

In #2445 I resolved the issue by adding to fincl3 the setting hist_avgflag_pertape(3) = 'I' with approval from @ekluzek.

General bug information

CTSM version you are using: preparing to merge ctsm5.3.062 (#2445)

Does this bug cause significantly incorrect results in the model's science? No

Configurations affected: LREPR* tests

[slevis-lmwg]

@billsacks @samsrabin
I assigned you only because you may be familiar with the line of python code that I referenced above and not because we know, yet, who will fix it.

[billsacks]

I see the problem, and might have a solution. More soon, but wanted to note this so others don't spend more time looking into it and duplicating effort....

[billsacks]

Okay, yeah, so the problem is a subtle one, and I think I have a fix if I'm understanding the bug correctly: The problem occurred when "GRAIN" appeared in two (or more) strings, where one was a substring of the other. For example, in LREPRSTRUCT_Ly1_P128x1.f10_f10_mg37.I1850Clm50BgcCrop.derecho_gnu.clm-ciso--clm-cropMonthOutput, before this replacement, one line contained 'GRAINN_TO_FOOD' and a later line contained (among other things) "'GRAINN_TO_FOOD_PERHARV', 'GRAINN_TO_FOOD_ANN'". This was problematic because the first replacement of GRAINN_TO_FOOD incorrectly led to replacements in the later strings as well.

#3314 fixes this issue.