Gracefully handle (ignore) null bytes in history lines

[owst]

Fixes #1789.

Readline is unable to add lines to its history that contain a null byte;
we should therefore avoid saving such lines to the history file, and
ignore any such lines that are already present in the file.

[kyrylo]

I wonder if we should substitute \0 with a blank instead? Like you said in #1789 you typed that by accident, right?

[owst]

When you say "substitute \0 with a blank" do you mean "we could delete \0s" or do you mean "we could replace each \0 with a (a space)"? I think you might mean the latter, since in #1789 I said I typed ctrl-space instead of space to generate \0.

I see some issues:

1. In general a null byte bears no relation to a space - ctrl-space is one keypress sequence to type a literal NULL byte, for me on Linux, the sequence ctrl-v, ctrl-shift-' also enters a NULL byte, and there's likely other ways too
2. There's no guarantee that the NULL byte appears in a position where adding a space is syntactically valid (e.g. x\0y = 1)
3. The simple symmetry between "don't save invalid lines" and "don't load existing invalid lines" would be broken

What do you think? I think the behaviour as-is in this PR is simple and likely to be most-often correct vs attempting to modify invalid lines

[kyrylo]

I think you might mean the latter [...]

That's correct. This is what I meant.

Thanks for the analysis, I agree with this decision. I was just wondering about the "what-if" but your points sound very convincing.