extraction: refine img src url and fix table extraction bugs

[unsleepy22]

1. refine img src url, allow for absolute and relative paths
2. fix bugs in table extraction: 1) the last row with colspan is not well formatted in markdown/txt; 2) remove newlines in table cell markdown in some cases

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.27%. Comparing base (76200b7) to head (0be7633).
Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #762   +/-   ##
=======================================
  Coverage   99.26%   99.27%           
=======================================
  Files          21       21           
  Lines        3559     3572   +13     
=======================================
+ Hits         3533     3546   +13     
  Misses         26       26           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[adbar]

Great work, the tests show that you definitely improve table processing with this PR.

I just have a few comments to make the code more readable and easier to maintain.

[unsleepy22]

Great work, the tests show that you definitely improve table processing with this PR.

I just have a few comments to make the code more readable and easier to maintain.

All fixed, would you take a look again?

[unsleepy22]

Great work, the tests show that you definitely improve table processing with this PR.

I just have a few comments to make the code more readable and easier to maintain.

There could be subtle bugs in table extraction as tables sometimes get really complicated, luckily I may encounter enough cases during my work (collecting LLM corpus), so hopefully bugs converge gradually and we may set up enough unit tests or real world test cases.