Memory improvements when double checking spelling corrections

The double checking of output that occurs after the initial strings.Replacer is not so good.

[x] native string concatenation (slow)
[x] buffers not reallocated
[x] buffers between diffLine and recheckLine not shared
[x] diffs founds are done in diffLine and recheckLine and copied
[x] fast case calls a function, only to return immediately
[x] use of string.Split on newlines (over potentially large file), instead of iteration TBD if this is good or bad.
[x] DiffLine is a public function for unknown reasons

client9 / misspell