Speedup pow_of_ten/1

[Yozhig]

This PR speedups decimal:add/2 and few other functions with usage of case expression for pow_of_ten calculations for exponents up to 100 (should be enough for most real-world cases).

How to test:

• checkout a commit with a bench (e386e10e16255bfa9f717ffe36210b5171d06c1a)
• create a baseline with rebar3 bench --save-baseline upstream
• checkout a commit with an optimization (41b0ee87d942952fb309a6580eb6bb01ce7dba94)
• bench and compare with rebar3 bench --baseline upstream

My results for OTP 26:

===> Testing bench_decimal:bench_add()
===> Stats for wall_time
Min:                 15.27μs (- 5394.00ns / 26.1%)
25 percentile:       15.49μs (- 5533.00ns / 26.3%)
Median:              15.62μs (- 5588.00ns / 26.3%)
75 percentile:       15.73μs (- 5635.00ns / 26.4%)
Max:                 15.91μs (- 7458.00ns / 31.9%)
Outliers:         Lo: 0; Hi: 0; Sum: 0
Outlier variance: 9.90000e-3 (unaffected)
> Bootstrapped
Mean:                15.62μs (- 5603.00ns / 26.4%)
Std deviation:      163.00ns
> Relative
Difference at 95.0 confidence
 -5603.00ns ±    74.00ns
   -26.40%  ±  0.35%
 (Student's t, pooled s = 268.668)

and OTP 24:

===> Testing bench_decimal:bench_add()
===> Stats for wall_time
Min:                 16.24μs (- 5275.00ns / 24.5%)
25 percentile:       16.50μs (- 5897.00ns / 26.3%)
Median:              16.67μs (- 5921.00ns / 26.2%)
75 percentile:       16.90μs (- 5844.00ns / 25.7%)
Max:                 17.08μs (- 6121.00ns / 26.4%)
Outliers:         Lo: 0; Hi: 0; Sum: 0
Outlier variance: 6.60675e-2 (slight)
> Bootstrapped
Mean:                16.69μs (- 5815.00ns / 25.8%)
Std deviation:      233.00ns
> Relative
Difference at 95.0 confidence
-5815.00ns ±    83.00ns
  -25.84%  ±  0.37%
(Student's t, pooled s = 300.593)