Create a speed optimized Preset

[npatki]

Problem Description

After adding the framework in #715, we should add a configuration that allows the user to optimize a tabular model for speed.

Expected behavior

A speed optimized tabular model will:

• Use the GaussianCopula
• Not round numerical data
• Enforce min/max values on numerical data
• Use the CategoricalTransformer (to be renamed: FrequencyEncoder) on any categorical columns
• Do not model null values. Instead, randomly generate null values based on observed proportions
• Default all univariate distributions to gaussian instead of performing a search

API

The TabularPreset will accept 2 parameters:

• (required) optimize_for, the name of the preset. We'll add "SPEED"
• metadata a dictionary or TableMetadata object, similar to tabular models today

It will print out hard-coded info about the configuration & benchmarks.

>>> from sdv.lite import TabularPreset

# the PresetModel will not allow too many extra settings, making it simple to use
>>> model = TabularPreset(optimize_for='SPEED', metadata=my_metadata)
Info: This config optimizes the modeling speed above all else.

Your exact runtime is dependent on the data. Benchmarks:
100K rows and 100 columns may take around 1 minute.
1M rows and 250 columns may take around 30 minutes.

# these work like any other tabular model
>>> model.fit(my_data)
>>> model.sample(my_data)

Edge Cases

If the user does not pass in metadata, we should still run the model, but we should provide a warning:

>>> from sdv.lite import TabularPreset
>>> model = TabularPreset(optimize_for='SPEED')
Warning: No metadata provided. Metadata will be automatically detected from your data. This process may
not be accurate. We recommend writing metadata to ensure correct data handling.

Info: This config optimizes the modeling speed above all else.

Your exact runtime is dependent on the data. Benchmarks:
100K rows and 100 columns may take around 1 minute.
1M rows and 250 columns may take around 30 minutes.

If the user does not pass in the required optimization parameter, throw an error.

>>> from sdv.lite import TabularPreset
>>> model = TabularPreset()
Error: You must provide the name of a preset using the 'optimize_for' parameter. Use 
'TabularPreset.list_available_presets()' to browse through the options.