Implement charset conversion on recording about tlog HOT 1 OPEN

This problem is similar to #10 in that converting anything in the recorded data would make the recording inexact and would greatly complicate recovery of binary data.

OTOH, we need to have the text in UTF-8 so that ElasticSearch can index and search it. So, to have both, i.e. to have ElasticSearch index the data and preserve the recorded data exactly, we either should limit the terminal charset to UTF-8, or log both the binary recording (e.g. in base64) and cleaned-up and converted text recording.

Perhaps the latter is what we should do. The latter will also require recording the original terminal charset.

In the meantime we can stay with the existing format, limit the encoding to UTF-8, add format version according to #15 and later implement recording both.

from tlog.