Refine fill forward strategy to handle issue with summary table

[prakaa]

Previously, the fill forward strategy for multi-header rows meant that single header columns after multi-headers would be mislabelled (see issue below for an example). This seems to only occur in the summary tables

To fix this, fill forward is manually implemented to:

1. Accept the intermediate header row and the preceding header row
2. If a value in the intermediate header row is nan, then:
   a) If value in corresponding column of the preceding header row is nan, then forward fill
   b) else leave as nan

This should work for 3+ header rows, but AEMO only seems to have at most 3 header rows at the moment

Ran all tests and produced all output - note that all changes are only new separator (this is branched off that PR) and removing wrong merged headers in summary files. So for easiest comparison, diff these changes with those of the new separator branch

Closes #23

[codecov]

Codecov Report

Attention: Patch coverage is 91.66667% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/isp_workbook_parser/read_table.py	90.90%	0 Missing and 1 partial ⚠️

Files with missing lines	Coverage Δ
src/isp_workbook_parser/config_model.py	100.00% <ø> (ø)
src/isp_workbook_parser/parser.py	84.66% <100.00%> (+4.00%)	⬆️
src/isp_workbook_parser/read_table.py	98.14% <90.90%> (-1.86%)	⬇️

[nick-gorman]

I've inspected how the functionality works on a couple of tables in summary mapping and it appears to be working properly.