fix(en-tn): verbalize plural time unit abbreviations (mins/hrs/secs/kgs/wks/mos/yrs)

[JMak-Security]

Supersedes #479, which the stale bot closed while CI was green and the PR was mergeable.

GitHub will not let me reopen #479: rebasing the branch onto current main required a force-push, and reopening is refused after a force-push (state cannot be changed ... the branch was force-pushed or recreated). So this is the same fix, rebased, with the regression coverage that was missing from the original revision, and a fresh PR to go with it.

What this does

en/data/measure/unit.tsv maps the singular time abbreviations min, sec, hr to their spoken forms, but not the plural written forms mins, hrs, secs. Since measure normalization only fires when the written token is recognised as a unit, the plural forms pass through unverbalized:

'5 mins'   -> 'five mins'          # should be 'five minutes'
'2 hrs'    -> 'two HRS'            # should be 'two hours'
'2-3 mins' -> 'two - three mins'   # should be 'two to three minutes'

The range case is the more visible failure: the measure range rule cannot fire either, so the hyphen from 2-3 is left in the output, which TTS renders as silence/a pause.

Scope

Limited to the units that take a colloquial s plural: min/mins, sec/secs, hr/hrs, kg/kgs, wk/wks, mo/mos, yr/yrs. SI symbols (km, mm, GB, kW, …) are deliberately untouched, as those are not written that way. kgs mirrors the already-existing lbs -> pound precedent.

The scope question raised in review on #479 is resolved in favour of exactly this set.

Verification

Run with python 3.10 + pynini==2.1.6.post1 on both trees:

pytest tests/nemo_text_processing/en/test_measure.py --cpu

input	main (before)	this PR (after)
5 mins	five mins	five minutes
2 hrs	two HRS	two hours
30 secs	thirty secs	thirty seconds
3 kgs	three kgs	three kilograms
2 wks	two wks	two weeks
6 mos	six mos	six months
5 yrs	five yrs	five years
2-3 mins	two - three mins	two to three minutes
5 min	five minutes	five minutes (unchanged)
1 sec	one second	one second (unchanged)

• main: 253 passed
• this PR: 268 passed (253 pre-existing + 15 new), 0 failed

15 cases were added to en/data_text_normalization/test_cases_measure.txt covering all seven plural units plus singular controls, so this cannot silently regress. The original revision changed data only, with no test exercising the behaviour, which is likely why it kept getting passed over.

Changes

1. fix(en-tn): verbalize plural time unit abbreviations (mins/hrs/secs)
2. fix(en-tn): add kgs/wks/mos/yrs plural unit abbreviations
3. test(en-tn): add regression cases for plural unit abbreviations

Could someone take a look when convenient? And if en TN fixes are no longer landing in this repo, I would appreciate a pointer to where they should go instead — happy to re-target.