Is it possible to transform data using a dynamic list of column names in KQL?

I have a Kusto table that contains compressed data, along with an array of strings representing the names of the columns containing said compressed data. The data in this array cannot be known prior to running the query, and can differ per record. I need to decompress the data in these columns (via gzip-base64-decompress). I'm hoping to go from something like the following input_table to output_table.

input_table:

output_table:

So far, I am able to obtain a table containing a row per column name containing data in need of decompression.

input_table
| extend intermediate = parse_json(compressed_properties)
| mv-expand ColumnToCompress = intermediate
| project ColumnToCompress

From here, I'm not sure how to proceed.

>I'm hoping to go from something like the following inpute_table to output_table.

I taken an input table like below ( and i have taken array with 2 values which is static list):

Now I have decompressed it using below kql query:

let inp = datatable(compressed_properties: dynamic, foo: string, bar: string)
[
  '["foo", "bar"]', 'H4sIAAAAAAAA/wEHAPj/Uml0aHdpa3r2/MsHAAAA', 'H4sIAAAAAAAA/wEFAPr/Ym9qamFwqHKmBQAAAA=='
];
inp
| extend foo = gzip_decompress_from_base64_string(foo)
| extend bar = gzip_decompress_from_base64_string(bar)

Fiddle.

Try to follow above query and integrate according to your requirement(if you are using many elements in array, you need to find a way to loop through it and then use extend).