[metif12]

tail: make both test files run again

Neither test file in this directory had been running. One did not compile, and
the other compared an empty string against its expected output five times.

string_to_i64_test.v:24  error: to propagate a Result, the call must also
                                return a Result type

string_to_i64('1')! == 1 used to unwrap. ! now means propagate, which needs
the enclosing function to return a Result, so the file has not compiled since
that change and v test . has been reporting it as a runtime failure with a zero
runtime. It is the only file in this repository that wrote it that way. Two
helpers, converted() and rejected(), replace the twenty-odd unwraps; the
assertions themselves are unchanged.

tail_test.v: all five tests, empty against expected

tail() is handed a closure to write through, and setup() built one that appended
to a &[]string pointing at a local in setup(). That local is gone by the time
tail() runs, so nothing was ever appended. Measured both obvious spellings on
current V:

closure appending to a captured `mut []string`   reads back empty
closure appending through a `&[]string`          reads back empty, and
                                                    crashes in array__push

A pointer to a heap struct does share, so the sink is a &Sink with a mut
field and the closures capture that. tail itself was never wrong here: tail -n 2 test.txt prints exactly what the test expected, byte for byte, and so does GNU.

One more thing, because make test builds with -W and a deprecation is an error
there: os.execute('ps ${args.pid}') is now os.exec(['ps', args.pid]). There is
no shell in it to lose, and -Wimpure-v -W now passes for this directory.

Tests

v test ./src/tail is 2 of 2 and so is v -Wimpure-v -W test ./src/tail.
v fmt -verify src/tail passes.

The whole suite is still 17 of 49 files failing on Windows; this takes it to 16.

[JalonSolov]

Conflicts.