fix: make col_sample min equals to 1

[HEGAB7]

Improved Column Sampling for Small Datasets

In our current implementation, col_sample is treated as a percentage. When the number of columns is small and col_sample is also small, calculating the sampled columns with:

col_size = int(self.col_sample * X.shape[1])

can result in col_size being 0.

For example:
if there are 4 columns and col_sample = 0.1, then int(0.1 * 4) yields 0, which subsequently causes an error during model generation due to having 0 input features.

This pull request updates the logic to ensure that at least one column is always selected when col_sample is positive by using:

col_size = max(1, int(self.col_sample * X.shape[1])) if self.col_sample > 0 else 0

This change prevents errors in cases with a small number of features, improving the robustness of the model creation process.

[ryan-wolbeck]

Looks good to me, thanks for the contribution!