Chunk calls to DBT based on string length

[willbowditch]

Fixes issue https://github.com/elementary-data/elementary/issues/1353

Bug

In some scenarios the argument list Elementary calls DBT with when uploading source freshness exceeds the limit of the operating system.

This is because the chunking of calls to DBT is currently based on the number of sources and not the length of the arguments. So some combination of sources results in longer argument lists than others.

This PR changes the chunking to use the character length and adds a buffer to allow for a bit of variation.

Testing

I've tested this locally by patching elementary in our deployment and confirming the results are uploaded.