seq_linter() could optionally lint all 1:<non-literal> usage

[MichaelChirico]

I think that a large portion of 1:<non-literal> usages should be replaced by seq_len(<non-literal>). In practice it's often as simple as:

size <- length(x)
for (i in 1:size) { }

1:10 should still be allowed as seq_len(10) is probably overkill.

I'm tempted to turn this on by default, but I'm OK with a more conservative approach to leave this under an option that's off by default.

[AshesITR]

SGTM. Regarding defaults, I'm also tempted.
I think we should check how many new lints this would produce on a random CRAN sample and decide based on the data.