[ENH] CSV Import: guess data types

[PrimozGodec]

Issue

Implements #4794

Description of changes

Implemented guessing strategy for CSV import which should match the strategy in io_utils.guess_data_type.

This PR adds another iteration over the columns. For each column, it checks for the data type according to data. Complex operations here are:

• unique: Pandas unique is based on a hash table - less complex than Numpy's one
• casting times to date-time

@ales-erjavec is this kind of guessing acceptable or would decrease the performance of the widget too much?

TODO

• Tests

Includes

• Code changes
• Tests
• Documentation

[codecov]

Codecov Report

Merging #4838 into master will increase coverage by 0.05%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #4838      +/-   ##
==========================================
+ Coverage   84.01%   84.07%   +0.05%     
==========================================
  Files         281      277       -4     
  Lines       56901    56487     -414     
==========================================
- Hits        47804    47490     -314     
+ Misses       9097     8997     -100     

[ales-erjavec]

is this kind of guessing acceptable or would decrease the performance of the widget too much?

I think it is ok.

[PrimozGodec]

Ok. Then I will write a few tests before it is merged.