Update Document id generation algorithm

[vblagoje]

During implementation of Azure converter issue one of the unit tests that worked on 1.x branch kept failing. The unit test istest_azure_converter_with_multicolumn_header_table and it is available on this PR

Upon further investigation @sjrl and I have traced this issue to stem from how we calculate Document id. A possible solution might be to "allow duplicate column names. Looking at the example in the PDF, it is very common to have tables in things like financial reports to have multi column headers. And I think the best way to represent that in a dataframe is to have duplicate column names. I think it would be better to update the call to the to_json() method to work with dataframes that have duplicate column names."

Not sure how deep consequences of this change would be but a PR resolving the Azure converter issue is blocked by this issue.

[vblagoje]

This issue is blocking:

• Feature parity for AzureConverter v1.9 in v2.0 + Overall Table Extraction Options #6468

cc @sjrl @silvanocerza @masci