Don't marshal empty byte or uint8 slice as null

[nikhita]

Fixes #272

[]byte or []uint8 are encoded as base-64 encoded strings. So, only nil slices of these types should get marshalled as null, and non-nil empty slices should get marshalled as "".

This restores compatibility with the standard library.

I verified that this patch works locally. Even though tests for these exist in https://github.com/json-iterator/go/blob/master/type_tests/slice_test.go, I couldn't reproduce a case where the test case was fuzzed to an empty []byte or []uint8 slice... 😕

/cc @taowen @thockin
fyi @liggitt @dims @neolit123

[nikhita]

cc @sttts

[dims]

@nikhita i was playing with this snippet - https://play.golang.org/p/IOXp4klmNvU

before applying your patch, both "case 1" and "case 2" end up printing marshal : "null"
after applying your patch, "case 1" prints marshal : "null" and "case 2" prints marshal : "\"\""

does this help?

[nikhita]

Updated with tests. Turned out that there are special tests (apart from the normal fuzzing of types) for encoding byte arrays.

[codecov]

Codecov Report

Merging #371 into master will increase coverage by 0.12%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #371      +/-   ##
==========================================
+ Coverage   81.69%   81.81%   +0.12%     
==========================================
  Files          41       41              
  Lines        5020     5021       +1     
==========================================
+ Hits         4101     4108       +7     
+ Misses        798      792       -6     
  Partials      121      121

Impacted Files	Coverage Δ
reflect_native.go	88.16% <100%> (+0.04%)	⬆️
reflect_struct_decoder.go	49.16% <0%> (+0.83%)	⬆️

---

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 08047c1...fb5614a. Read the comment docs.

[dims]

@nikhita do we have to add the test for the var mySlice1 []byte // Case 1? where it's marshalled to null?

[nikhita]

@dims Done.

[dims]

LGTM. Thanks @nikhita

[thockin]

LGTM. I try to be judicious in my authority on this repo, but this seems like an important edge case.