[enhancement] NSDL and CDSL Parsers

[AmanJagotra]

Currently parsing of holdings only should suffice, we can target transactions later on as per requirement/demand

• Parsing should be as accurate as possible with all required information, if any caveats, please mention.
• Consistent with current code structure

We might have to explore PyMuPDF Table, if possible, to get the required accuracy

[AmanJagotra]

PyMuPDF Tables are quite slow as compared to the textual approach currently being used as when grid lines are not defined (in case of NSDL), vector graphics are being used to detect edges.
Will be glad to move to tables approach once we find an alternative.