Fix section_text last_paragraph start index off by separator length

[Chessing234]

Bug

section_parser.section_text returns a section_idx list documented as "list of start indices of the text in the section" — offsets back into the original report string. When the report has no impression/findings section and the final section contains '\n \n', the function synthesises a 'last_paragraph' section by splitting on that separator:

if '\n \n' in sections[-1]:
    sections.append('\n \n'.join(sections[-1].split('\n \n')[1:]))
    sections[-2] = sections[-2].split('\n \n')[0]
    section_names.append('last_paragraph')
    section_idx.append(section_idx[-1] + len(sections[-2]))

The computed offset for the new 'last_paragraph' section is section_idx[-1] + len(sections[-2]), but sections[-2] at that point holds only the first part of the split — the original section up to (but not including) the '\n \n' separator. str.split removes the separator, so the second part (which becomes the new 'last_paragraph') actually starts len('\n \n') (3) characters later in the original text.

Root cause

The offset math omits the length of the separator that split() consumed. For every other entry in section_idx, the value is the end of the section-header regex match (s.end()) — i.e. already past the :\s separator — so using raw offsets relative to the original text is the invariant callers rely on. This branch breaks that invariant.

Why the fix is correct

Adding len('\n \n') puts the offset at the first character of the last paragraph in the original text, matching what text[section_idx[i]:] gives for every other produced index:

-            section_idx.append(section_idx[-1] + len(sections[-2]))
+            section_idx.append(section_idx[-1] + len(sections[-2]) + len('\n \n'))

• sections[-2] is the first part (unchanged by this commit).
• section_idx[-1] is still the original section's start index at the point of append (not yet mutated).
• len('\n \n') is the separator length that str.split('\n \n') drops.

No other behavioural change: the returned sections list is identical, section_names is identical, only the previously-incorrect section_idx[-1] for the last_paragraph entry moves forward by 3 to the correct position.