Add blob type for arbitrary binary data

[jachris]

We currently use the string type both for printable strings and for arbitrary binary data. However, this violates Arrow's UTF-8 requirements and leads to several other issues down the road (for example: it's not possible to print arbitrary blobs as JSON strings). Thus, this PR introduces a new blob type, that is designed to handle arbitrary binary data.

Internally, this binary data is directly stored as an arbitrary sequence of bytes (using Arrow's BinaryType for table slices). However, when the data enters or leaves Tenzir, we must have some special handling, depending on the format. For example, assume we want to print the blob b"\x00\xFF" as JSON. This blob cannot be represented as an escaped string. Thus, we base64 encode it, and print "AP8=". When parsing JSON, we cannot assume that all base64-valid strings are binary payloads. However, when a schema with a blob type is given, we base64 decode the blob to get b"\x00\xFF" back. Other formats might use different strategies to encode and decode blobs.

[dominiklohmann]

This is really nice, thanks for hacking this together so quickly. I have some questions on the code and some answers to questions you left in the code. I'll play around with this in practice some more to test it further and will then approve it afterwards.

[dominiklohmann]

This works well enough in practice to be approved. I think once the outstanding discussions are resolved we can merge this and be done with it.

[netantho]

This PR solves an issue of stalled pipeline when importing tshark data in (nd)JSON, thanks!