Use reflect.Value.IsZero for omitempty checks

[NathanBaulch]

Struct fields of type time.Time and netip.Addr are always omitted when using omitempty because the current isEmptyValue function can only check structs with exported fields. See #955 and #988.

Replacing the current isEmptyValue function with reflect.Value.IsZero() is a very simple and low risk fix that should make omitempty useful for many more types.

---

goos: linux
goarch: amd64
pkg: github.com/pelletier/go-toml/v2/benchmark
cpu: Intel(R) Core(TM) Ultra 9 275HX
                                   │ old_bench_10.txt │          new_bench_10.txt          │
                                   │      sec/op      │   sec/op     vs base               │
UnmarshalDataset/config-24                11.95m ± 3%   12.28m ± 3%       ~ (p=0.105 n=10)
UnmarshalDataset/canada-24                50.90m ± 6%   46.93m ± 3%  -7.80% (p=0.002 n=10)
UnmarshalDataset/citm_catalog-24          21.77m ± 4%   21.27m ± 1%  -2.31% (p=0.035 n=10)
UnmarshalDataset/twitter-24               7.624m ± 6%   8.092m ± 4%  +6.15% (p=0.029 n=10)
UnmarshalDataset/code-24                  51.21m ± 6%   50.42m ± 4%       ~ (p=0.165 n=10)
UnmarshalDataset/example-24               111.4µ ± 3%   115.9µ ± 5%       ~ (p=0.089 n=10)
Unmarshal/SimpleDocument/struct-24        354.5n ± 3%   380.2n ± 3%  +7.25% (p=0.000 n=10)
Unmarshal/SimpleDocument/map-24           494.8n ± 4%   517.0n ± 4%  +4.50% (p=0.029 n=10)
Unmarshal/ReferenceFile/struct-24         27.74µ ± 5%   29.04µ ± 5%  +4.68% (p=0.035 n=10)
Unmarshal/ReferenceFile/map-24            42.84µ ± 5%   42.85µ ± 4%       ~ (p=0.853 n=10)
Unmarshal/HugoFrontMatter-24              7.200µ ± 3%   7.374µ ± 3%       ~ (p=0.143 n=10)
Marshal/SimpleDocument/struct-24          258.1n ± 6%   263.9n ± 3%       ~ (p=0.271 n=10)
Marshal/SimpleDocument/map-24             311.6n ± 2%   313.6n ± 3%       ~ (p=0.565 n=10)
Marshal/ReferenceFile/struct-24           21.15µ ± 3%   21.56µ ± 1%  +1.93% (p=0.011 n=10)
Marshal/ReferenceFile/map-24              25.35µ ± 4%   25.54µ ± 4%       ~ (p=0.247 n=10)
Marshal/HugoFrontMatter-24                4.782µ ± 4%   4.859µ ± 3%       ~ (p=0.224 n=10)
geomean                                   67.32µ        68.44µ       +1.65%

                                   │ old_bench_10.txt │          new_bench_10.txt           │
                                   │       B/s        │     B/s       vs base               │
UnmarshalDataset/config-24               83.70Mi ± 3%   81.45Mi ± 3%       ~ (p=0.105 n=10)
UnmarshalDataset/canada-24               41.25Mi ± 6%   44.73Mi ± 3%  +8.45% (p=0.002 n=10)
UnmarshalDataset/citm_catalog-24         24.45Mi ± 4%   25.02Mi ± 1%  +2.36% (p=0.034 n=10)
UnmarshalDataset/twitter-24              55.28Mi ± 6%   52.08Mi ± 4%  -5.80% (p=0.029 n=10)
UnmarshalDataset/code-24                 49.99Mi ± 6%   50.77Mi ± 4%       ~ (p=0.165 n=10)
UnmarshalDataset/example-24              69.36Mi ± 3%   66.66Mi ± 5%       ~ (p=0.089 n=10)
Unmarshal/SimpleDocument/struct-24       29.59Mi ± 3%   27.59Mi ± 3%  -6.75% (p=0.000 n=10)
Unmarshal/SimpleDocument/map-24          21.20Mi ± 4%   20.29Mi ± 4%  -4.30% (p=0.029 n=10)
Unmarshal/ReferenceFile/struct-24        180.2Mi ± 5%   172.1Mi ± 5%  -4.45% (p=0.035 n=10)
Unmarshal/ReferenceFile/map-24           116.7Mi ± 5%   116.7Mi ± 4%       ~ (p=0.853 n=10)
Unmarshal/HugoFrontMatter-24             72.33Mi ± 3%   70.61Mi ± 3%       ~ (p=0.143 n=10)
Marshal/SimpleDocument/struct-24         44.33Mi ± 5%   43.36Mi ± 3%       ~ (p=0.271 n=10)
Marshal/SimpleDocument/map-24            36.73Mi ± 2%   36.50Mi ± 3%       ~ (p=0.542 n=10)
Marshal/ReferenceFile/struct-24          92.75Mi ± 3%   90.99Mi ± 1%  -1.90% (p=0.009 n=10)
Marshal/ReferenceFile/map-24             75.47Mi ± 4%   74.90Mi ± 4%       ~ (p=0.247 n=10)
Marshal/HugoFrontMatter-24               104.3Mi ± 4%   102.7Mi ± 3%       ~ (p=0.247 n=10)
geomean                                  58.41Mi        57.46Mi       -1.62%

All the other measurements were equal.

[pelletier]

My apologies for the very long response delay.

I think we should follow encoding/json and support a omitzero tag to not break the current behavior and follow encoding/json's behavior.

[pelletier]

Closing because #998 is merged now!