Fix for CPI_offset logic

[hdoupe]

When the CPI_offset parameter is updated in 2018 or later, the values from prior years are re-applied to the inflation rates. Here's an example demonstrating the behavior:

import numpy as np

from taxcalc import Policy


p1 = Policy()
p2 = Policy()
p2.implement_reform({"CPI_offset": {2022: -0.005}})

p1_rates = p1.inflation_rates()
p2_rates = p2.inflation_rates()

for year in range(2013, 2030):
    print(
        year,
        p1_rates[year - 2013],
        p2_rates[year - 2013],
        np.round(p1_rates[year - 2013] - p2_rates[year - 2013], 4)
    )

# output:
2013 0.0148 0.0148 0.0
2014 0.0159 0.0159 0.0
2015 0.0012 0.0012 0.0
2016 0.0127 0.0127 0.0
2017 0.0187 0.0162 0.0025 # p1 is 0.0025 greater than p2
2018 0.0224 0.0199 0.0025 # even though they should be equal
2019 0.0186 0.0161 0.0025
2020 0.0233 0.0208 0.0025
2021 0.0229 0.0204 0.0025
2022 0.0228 0.0178 0.005
2023 0.0221 0.0171 0.005
2024 0.0211 0.0161 0.005
2025 0.0209 0.0159 0.005
2026 0.0211 0.0161 0.005
2027 0.0208 0.0158 0.005
2028 0.021 0.016 0.005
2029 0.021 0.016 0.005

This PR only applies the offset starting in the first year that CPI_offset is modified in the revision.

[codecov]

Codecov Report

Merging #2381 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff           @@
##           master   #2381   +/-   ##
======================================
  Coverage     100%    100%           
======================================
  Files          13      13           
  Lines        2748    2749    +1     
======================================
+ Hits         2748    2749    +1

Impacted Files	Coverage Δ
taxcalc/parameters.py	100% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 05702c0...c48a4b0. Read the comment docs.

[MattHJensen]

Thanks @hdoupe. I was able to replicate the bug and fix locally. Merging.