Improve TableFileParser column-type deduction for missing values

[pfleck]

A DateTime-column, where the first value is missing, is wrongly interpreted as string-column.

Column-type deduction of the TableFileParser currently works in the following way:

1. The initial type of a column based on the values of the first data-row.
2. If the first value is missing, the column is a double-column.
3. If a new value is not compatible with the current column-type, the column is converted to a string-column.

In case of a DateTime with the first value missing, the parser first decides it is a double-column and later converts it to a string-column when the first DateTime-value appears.

The TableFileParser should be improved, so that the column-type deduction is deferred until the first non-missing value appears.

[pfleck]

Added Patch that treats columns of unknown types as List<object> and performs conversion to a specific list-type when the first non-missing value occurs for that column.

[gkronber]

Reviewed the patch. Looks good.

[pfleck]

r15513 Fixed problem by temporarily using a List<object> to represent an unknown column-type until the type is known.

[pfleck]

r15537: merged r15513 to stable