arrow

ARROW-9992: [C++][Python] Refactor python to arrow conversions based on a reusable conversion API

### Targets of the refactoring:
- PythonToArrow converters based on a common API
- PyBytesView to use `Result` return values and contain `is_utf8` flag
- PyConversionOptions is now available from all converters so we can honor its flags

### Fixes

- ARROW-9993 [Python] Tzinfo - string roundtrip fails on pytz.StaticTzInfo objects
- ARROW-9994 [C++][Python] Auto chunking nested array containing binary-like fields result malformed output
- ARROW-9996 [C++] Dictionary is unset when calling DictionaryArray.GetScalar for null values
- ~ARROW-9997 [Python] StructScalar.as_py() fails if the type has duplicate field names~
- ARROW-9999 [Python] Support constructing dictionary array directly through pa.array()
- ARROW-10000 [C++][Python] Support constructing StructArray from list of key-value pairs
- ARROW-9593 [Python] Add custom pickle reducers for DictionaryScalar
- ARROW-6281 [Python] Produce chunked arrays for nested types in pyarrow.array
- ARROW-2367 [Python] ListArray has trouble with sizes greater than kMaximumCapacity
- ARROW-9976: [Python] ArrowCapacityError when doing Table.from_pandas with large dataframe

### Backward incompatibility

~~Since a struct type can contain duplicated field names we cannot return a struct scalar as a mapping, so I had to change the `.as_py()` representation to return with a list of key-value pairs.~~

### TODOs:
- [x] ensure that the large memory tests are passing
- [x] benchmark and check binary size again

### Library size

Before:
```
 12M Sep 25 15:05 libarrow.200.0.0.dylib
2.7M Sep 25 15:07 libarrow_python.200.0.0.dylib
```

After:
```
 12M Sep 25 15:46 libarrow.200.0.0.dylib
2.1M Sep 25 15:50 libarrow_python.200.0.0.dylib
```

### Benchmarks

Executed the following ASV benchmark:

```bash
asv continuous --bench convert_builtins master py2ar --no-only-changed --split
```

After some optimization:

```
Benchmarks that have improved:

       before           after         ratio
     [f358a29b]       [18d1c052]
     <master>         <py2ar>
-     2.78±0.03ms      2.45±0.03ms     0.88  convert_builtins.ConvertPyListToArray.time_convert('bool')
-     3.59±0.01ms      3.12±0.02ms     0.87  convert_builtins.ConvertPyListToArray.time_convert('int32')
-     3.37±0.01ms      2.73±0.01ms     0.81  convert_builtins.ConvertPyListToArray.time_convert('uint32')
-     3.74±0.02ms      3.03±0.01ms     0.81  convert_builtins.ConvertPyListToArray.time_convert('int64')
-     3.38±0.01ms      2.69±0.01ms     0.80  convert_builtins.ConvertPyListToArray.time_convert('uint64')
-     2.83±0.01ms      2.24±0.01ms     0.79  convert_builtins.ConvertPyListToArray.time_convert('float32')
-     3.92±0.02ms      2.99±0.02ms     0.76  convert_builtins.ConvertPyListToArray.time_convert('binary10')
-     14.1±0.04ms      8.89±0.05ms     0.63  convert_builtins.ConvertPyListToArray.time_convert('unicode')
-     5.60±0.01ms      3.24±0.03ms     0.58  convert_builtins.ConvertPyListToArray.time_convert('ascii')
-     5.37±0.02ms      2.91±0.04ms     0.54  convert_builtins.ConvertPyListToArray.time_convert('binary')

Benchmarks that have stayed the same:

       before           after         ratio
     [f358a29b]       [18d1c052]
     <master>         <py2ar>
      14.8±0.02ms       15.5±0.1ms     1.05  convert_builtins.ConvertPyListToArray.time_convert('decimal')
       16.4±0.7ms       15.1±0.6ms     0.92  convert_builtins.ConvertPyListToArray.time_convert('struct from tuples')
       34.4±0.3ms       31.5±0.4ms     0.92  convert_builtins.ConvertPyListToArray.time_convert('int64 list')
       16.7±0.7ms       15.1±0.6ms    ~0.91  convert_builtins.ConvertPyListToArray.time_convert('struct')
      2.42±0.02ms      2.05±0.03ms    ~0.85  convert_builtins.ConvertPyListToArray.time_convert('float64')
```

Closes #8088 from kszucs/py2ar

Authored-by: Krisztián Szűcs <szucs.krisztian@gmail.com>
Signed-off-by: Benjamin Kietzman <bengilgit@gmail.com>