fix: return only requested pages from get_page_content on Markdown

[akhilesharora]

get_page_content's docstring says '3,8' means pages 3 and 8 (pageindex/retrieve.py:111-119). PDFs honor that. Markdown doesn't: _get_md_page_content takes min(page_nums)/max(page_nums) and returns everything in between.

Same input, before:

md  pages="5,100" -> [5, 10, 50, 100]
pdf pages="5,100" -> [5, 100]

After:

md  pages="5,100" -> [5, 100]
pdf pages="5,100" -> [5, 100]

_parse_pages already returns a discrete sorted list, so the simplest fix is to match against set(page_nums) instead of [min..max]:

wanted = set(page_nums)
...
if ln in wanted and ln not in seen:

Range form ('5-7') still parses to [5,6,7] so it keeps working. Only the comma-list shape changes.

Added tests/test_retrieve_pages.py with four cases. Two of them fail on main and pass with the fix. The other two cover the range and single-page forms to make sure nothing regresses there.

$ python -m pytest tests/test_retrieve_pages.py -v
==================== 4 passed in 1.87s =====================================

Files touched:

• pageindex/retrieve.py: 4 lines of logic, plus a docstring update so it matches the new behavior.
• tests/test_retrieve_pages.py: new, self-contained, no LLM or PDF needed.

No PDF behavior change, no API surface change.

Closes #279