issue ValueError exception when header=False and label_column is not an int

[jfomhover]

When header=False, the value in label_column should always be an int, not a str (ex: name:Label). This will cause lightgbm train to fail with an exception Could not find label column X in data file.

Instead, we could catch these kind of invalid parameter and provide clarity on the exception in the python preceeding the lightgbm train call.