feat(arrow): Streamline Apache Arrow extension types

[candiduslynx]

Common

Use pointer receivers:

1. For Array types: to have all functions have the same receiver type (other functions do have pointer receivers, so updating String() method receiver is natural, see net.IPNet as a reference.
2. For Type implementations: we do return a pointer from NewXYZType function, so it's only natural to define the receiver as pointer as well.

Inet

InetBuilder

1. Now accepts *net.IPNet (pointer) in Append, UnsafeAppend & AppendValues
2. AppendValues now accounts for valid param

InetArray

1. String method uses %q format as mainstream Apache Arrow types
2. Added Value(i int) *net.IPNet function
3. GetOneForMarshal utilizes Value implementation

JSON

JSONBuilder

1. Marshaling with json.DisableHTMLEscape() option per Destination plugins receive escaped data #622
2. AppendValues now accounts for valid param
3. UnmarshalJSON implementation is streamlined with helper Unmarshal(dec *json.Decoder) error function (as mainstream Apache Arrow types)

JSONArray

1. String method uses %q format as mainstream Apache Arrow types
2. Added Value(i int) any function that will unmarshal the stored data & return the concrete type
3. Use json.UnmarshalNoEscape per Destination plugins receive escaped data #622
4. GetOneForMarshal now doesn't perform unmarshaling and, instead, returns the stored json.RawMessage

Mac

MacBuilder

1. AppendValues now accounts for valid param

MacArray

1. String method uses %q format as mainstream Apache Arrow types
2. Added Value(i int) net.HardwareAddr function

UUID

UUIDBuilder

1. AppendValues now accounts for valid param
2. Use uuid.Must func instead of panic in code

UUIDArray

1. String method uses %q format as mainstream Apache Arrow types
2. Added Value(i int) uuid.UUID func
3. Use uuid.Must func instead of panic in code

[github-actions]

⏱️ Benchmark results

Comparing with 1a8072f

• DefaultConcurrencyDFS-2 resources/s: 11,068 ⬆️ 2.50% increase vs. 1a8072f
• DefaultConcurrencyRoundRobin-2 resources/s: 12,190 ⬆️ 2.03% increase vs. 1a8072f
• Glob-2 ns/op: 173.7 ⬇️ 43.58% decrease vs. 1a8072f
• TablesWithChildrenDFS-2 resources/s: 29,793 ⬆️ 18.92% increase vs. 1a8072f
• TablesWithChildrenRoundRobin-2 resources/s: 28,855 ⬆️ 9.48% increase vs. 1a8072f
• TablesWithRateLimitingDFS-2 resources/s: 28.35 ⬇️ 0.53% decrease vs. 1a8072f
• TablesWithRateLimitingRoundRobin-2 resources/s: 842.1 ⬆️ 3.46% increase vs. 1a8072f
• BufferedScanner-2 ns/op: 9.368 ⬇️ 24.15% decrease vs. 1a8072f
• LogReader-2 ns/op: 30.47 ⬇️ 25.40% decrease vs. 1a8072f

[codecov]

Codecov Report

Patch coverage: 48.12% and project coverage change: +0.20 🎉

Comparison is base (1a8072f) 46.98% compared to head (af39af0) 47.19%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #823      +/-   ##
==========================================
+ Coverage   46.98%   47.19%   +0.20%     
==========================================
  Files          76       76              
  Lines        7872     7844      -28     
==========================================
+ Hits         3699     3702       +3     
+ Misses       3682     3642      -40     
- Partials      491      500       +9     

Impacted Files	Coverage Δ
schema/arrow.go	13.39% <0.00%> (ø)
types/inet.go	41.73% <40.54%> (-0.67%)	⬇️
types/uuid.go	43.75% <40.62%> (+3.62%)	⬆️
types/mac.go	45.73% <52.00%> (-3.12%)	⬇️
types/json.go	60.30% <58.06%> (+10.70%)	⬆️

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[yevgenypats]

Im not sure I understand the reasoning or the value of using pointer receivers in some of the extensions instead of regular params? Were there any performance issue? we turn them into string anyway and by using pointers we just introduce more risk of the data being changed inside while we don't really want that to happen so I can't see the value unless Im missing something.