Fix identifier storage in Alignment class

[thomashopf]

Currently memory usage is defined by longest identifier due to use of numpy for identifier storage, which can create a large overhead if one header is longer than others - but numpy functionality not that relevant on identifiers

Ideally, replace with pd.Series to keep slicing functionality while making use of better string memory management of pandas

@aaronkollasch

[thomashopf]

Also add an option to from_file method to split identifiers on first whitespace (off by default)