Make eawk accept a custom separator

[saolof]

One of the biggest pain points of using elvish as a shell in practice for
me has been parsing output from shell commands that return text.

Many CLIs are parseable by AWK. Elvish provides eawk to provide
some of the key tasks that AWK does without introducing a
separate DSL, relying solely on the expressiveness of lambdas.

In practice, most of the power of AWK is being able to change
the separator to parse the particular output of the command
you pipe into it (commas, pipe, spacing strictly longer than two spaces, etc etc).
So this commit adds an optional sep argument that takes a regex pattern.

(I also updated the dockerfile.)

[krader1961]

It's great seeing someone new take the time to create a pull-request. But you also need to update the unit test to verify the new functionality. See the TestEawk function in builtin_fn_str_test.go.

[saolof]

Yes, I am putting together a couple of examples. I think sample outputs from psql and from docker image ls should work?

[saolof]

@krader1961 Added the docker image ls test.

[xiaq]

Thanks for the contribution.

I find the name of the options (&sep and &posix) potentially confusing though. It's not clear that &sep should be a regular expression (instead of a literal string to pass to str:split) and that &posix refers to the syntax of the regular expression.

I think it's best to move the command into the re: module altogether and have it named re:awk instead - the option names becomes unambiguous this way. The eawk name can remain now as a deprecated alias and be removed later.

I'll merge this to a temporary branch which I'll integrate to master later.

Maybe a re:sed is also in order...