Fix local executor failing on output lines over 64KiB

[paskal]

Previously, Local.Run split the captured output with a default bufio.Scanner, so a command printing a single line longer than 64KiB failed with bufio.Scanner: token too long even though the command itself had succeeded. The remote executor has no such limit, since Remote.sshRun splits the buffered output with strings.SplitSeq, so the same playbook behaved differently depending on whether it ran locally or over ssh.

On master:

$ cat longline.sh
#!/bin/sh
awk 'BEGIN{printf "%100000s\n", ""}'

$ spot --local -p longline.yml -t localhost
[0] failed command "emit one long line" on host localhost:0 (localhost): can't run script on localhost:0: bufio.Scanner: token too long

After this change the same playbook completes. The output is fully buffered in memory by the time it is split, so the scanner's bounded reads bought nothing, and splitting the buffer in place returns a line of any length as it is. The splitting keeps bufio.ScanLines semantics, dropping a trailing \r and the empty element after the final newline, so nothing changes for existing playbooks; TestSplitOutputLines pins that down case by case.

[umputun]

not blocking, merging as is.

the new helper keeps blank lines, while Remote.sshRun (remote.go:361) and Dry.Run (dry.go:30) drop them with if line != "", so the three executor.Interface implementations now return different line shapes for the same stdout. git log -S SplitSeq shows the only commit that touched those loops is 7640456, a mechanical Split -> SplitSeq rewrite, so nothing decided this deliberately.

two ways to go. Leave it and keep per-executor behavior, which costs nothing. Or unify on splitOutputLines so all three return the same shape, which changes Remote output: commands.go:564 joins with "; " and would start picking up empty segments. That would be a separate PR.