Optimize default behavior if multiple responses found in HAR data

Currently, in `TextReader._parse_raw_data()`, if multiple SAML responses are found in a HAR file, we just use the most recently generated SAML response data in the HAR file.

It might be nice to do something more sophisticated, such as:
- Prompting the user to select a SAML response to analyze, giving them information about the time acquired and the destination URL
- Analyzing the destination URL to see if it matches the format for the Assertion Consumer Service URL. This could break if the user configured their ACS URL incorrectly in their identity provider.

Thinking out loud, it may be better to do what is suggeseted in #29 and write a small application to clean up the HAR data file. You could prompt the user to select a SAML response and it would write out the raw SAML data to a file or to `stdout`, which could then be fed/piped into SAML Reader.