ODS: optimize handling table:number-columns-repeated attribute

[dimastbk]

This PR improves the performance of parsing ODS file, which produced by some spreadsheets, like LibreOffice Calc or Google Sheet, which add table:number-columns-repeated with large value, about 16000. We can make pushing empty cell with table:number-columns-repeated attribute lazy. See also the same optimization for table:number-rows-repeated in #308.

• tests/issues.ods

  $ cargo benchcmp old new
  name        old ns/iter  new3 ns/iter  diff ns/iter   diff %   speedup 
  bench_ods   18,599,072   79,045         -18,520,027  -99.58%  x 235.30 
  bench_xls   32,034       33,291               1,257    3.92%    x 0.96 
  bench_xlsb  134,122      132,433             -1,689   -1.26%    x 1.01 
  bench_xlsx  206,179      207,691              1,512    0.73%    x 0.99

• tests/special_cells.ods (w/o table:number-columns-repeated)

  $ cargo benchcmp old_wo new_wo
  name        old_wo ns/iter  new_wo ns/iter  diff ns/iter  diff %  speedup 
  bench_ods   38,891          39,564                   673   1.73%   x 0.98 
  bench_xls   32,973          33,084                   111   0.34%   x 1.00 
  bench_xlsb  140,205         137,071               -3,134  -2.24%   x 1.02 
  bench_xlsx  206,299         213,853                7,554   3.66%   x 0.96

The new test add test of handling merged cell (table:covered-table-cell) before empty cell.

[tafia]

there seem to be some conflict

[tafia]

Thanks for the PR. Do you mind chcking the conflict? Thanks

[dimastbk]

@tafia rebased.

[tafia]

Thanks