Force to use the Spark's system default precision and scale when inferred data type contains DecimalType.

[ueshin]

The return type inference sometimes fails when the inferred type contains DecimalType to infer the precision or the scale smaller.

>>> from decimal import Decimal
>>> kdf = ks.DataFrame({'a': [Decimal('1.1'), Decimal('2.2'), Decimal('10.01')]})
>>> kdf
       a
0   1.10
1   2.20
2  10.01
>>> kdf.spark.print_schema()
root
 |-- a: decimal(4,2) (nullable = false)

>>> with ks.option_context("compute.shortcut_limit", 1):
...   kdf.transform(lambda x: x).spark.print_schema()
...
root
 |-- a: decimal(3,2) (nullable = true)

>>> with ks.option_context("compute.shortcut_limit", 1):
...   kdf.transform(lambda x: x)
...
Traceback (most recent call last):
...
pyarrow.lib.ArrowInvalid: Decimal type with precision 4 does not fit into precision inferred from first array element: 3
...

We should use the Spark's system default precision and scale instead to avoid as many such cases as possible:

>>> with ks.option_context("compute.shortcut_limit", 1):
...   kdf.transform(lambda x: x).spark.print_schema()
...
root
 |-- a: decimal(38,18) (nullable = true)

>>> with ks.option_context("compute.shortcut_limit", 1):
...   kdf.transform(lambda x: x)
...
                       a
0   1.100000000000000000
1   2.200000000000000000
2  10.010000000000000000

[itholic]

Everything looks clear to me except one nit.

[codecov-io]

Codecov Report

Merging #1904 (bcb403d) into master (44d45f2) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #1904   +/-   ##
=======================================
  Coverage   94.16%   94.17%           
=======================================
  Files          41       41           
  Lines        9975     9989   +14     
=======================================
+ Hits         9393     9407   +14     
  Misses        582      582           

Impacted Files	Coverage Δ
databricks/koalas/accessors.py	93.00% <100.00%> (ø)
databricks/koalas/frame.py	96.73% <100.00%> (ø)
databricks/koalas/groupby.py	91.48% <100.00%> (ø)
databricks/koalas/namespace.py	83.33% <100.00%> (+0.03%)	⬆️
databricks/koalas/spark/utils.py	100.00% <100.00%> (ø)

---

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 44d45f2...bcb403d. Read the comment docs.

[ueshin]

Thanks! merging.