Assuming columns is supposed to be an array of name/type pairs (added closing ]) and that there are supposed to be 9 of these pairs (added Comment), and that you have a properly formatted JSON string (added surrounding and closing braces), then you could do something like this
| makeresults
| eval _raw="{\"properties\": {\"nextLink\": null,
\"columns\": [
{\"name\": \"Cost\", \"type\": \"Number\"},
{\"name\": \"Date\", \"type\": \"Number\"},
{\"name\": \"Charge\", \"type\": \"String\"},
{\"name\": \"Publisher\", \"type\": \"String\"},
{\"name\": \"Resource\", \"type\": \"String\"},
{\"name\": \"Resource\", \"type\": \"String\"},
{\"name\": \"Service\", \"type\": \"String\"},
{\"name\": \"Standard\", \"type\": \"String\"},
{\"name\": \"Comment\", \"type\": \"String\"}],
\"rows\": [
[2.06, 20210807, \"usage\", \"uuuu\", \"hhh\", \"gd\", \"bandwidth\", \"azy\", \"HHH\"],
[2.206, 20210807, \"usage\", \"uuuhhh\", \"ggg\", \"gd\", \"bandwidth\", \"new\", \"YYY\"]] }}"
| spath path="properties.columns{}.name" output=columnnames
| spath path="properties.rows{}{}" output=rows
| streamstats count as event
| mvexpand rows
| streamstats count as row by event
| eval index=(row-1)%mvcount(columnnames)
| eval name=mvindex(columnnames,index)
| eval {name}=rows
| eval row=floor((row-1)/mvcount(columnnames))
| fields - columnnames name index rows
| stats values(*) as * by row event
