Skip null categories when serializing.

[zbraniecki]

When we serialize plural rules, we can safely skip empty categories.

This reduces the size on the disk:

▶ dust -s -d2 -b ./json-old
  50B   ┌── manifest.json
 9.6K   │ ┌── ordinal@1
  22K   │ ├── cardinal@1
  32K   ├─┴ plurals
  32K ┌─┴ json-old

▶ dust -s -d2 -b ./json
  50B   ┌── manifest.json
 4.9K   │ ┌── ordinal@1
  13K   │ ├── cardinal@1
  18K   ├─┴ plurals
  18K ┌─┴ json

[coveralls]

Pull Request Test Coverage Report for Build ea0eb3ceba64bcc5b06ed696e1ac82942991ecc1-PR-251

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 78.951%

---

Totals
Change from base Build 7cc838c31053e52d660c690be850bfdfbe899cb0:	0.0%
Covered Lines:	2153
Relevant Lines:	2727

---

💛 - Coveralls

[zbraniecki]

I also see a modest improvement on the microbenchmark:

plurals/convert+select/fs
                        time:   [1.8451 us 1.8557 us 1.8702 us]
                        change: [-5.6253% -4.8749% -4.1690%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 13 outliers among 100 measurements (13.00%)
  1 (1.00%) low mild
  5 (5.00%) high mild
  7 (7.00%) high severe

plurals/select/fs       time:   [1.8258 us 1.8324 us 1.8392 us]
                        change: [-4.3913% -3.5502% -2.6586%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 11 outliers among 100 measurements (11.00%)
  1 (1.00%) low mild
  3 (3.00%) high mild
  7 (7.00%) high severe

plurals/construct/fs/10 time:   [383.39 us 385.23 us 387.16 us]
                        change: [-8.1226% -7.2399% -6.3760%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 11 outliers among 100 measurements (11.00%)
  2 (2.00%) low mild
  3 (3.00%) high mild
  6 (6.00%) high severe

[sffc]

Nice!!