fix StatLm duplicate formula arg when lm.args supplies formula

[jiyunson]

Hey Randall -- just a tiny edit! Thanks as always for all your awesome work on mosaic/ggformula/etc.

Summary

When lm.args = list(formula = y ~ poly(x, 2)) is passed to gf_lm() or stat_lm(), the call to do.call(stats::lm, c(base.args, lm.args)) fails with:

formal argument "formula" matched by multiple actual arguments

because base.args already contains formula = y ~ x (the default), and lm.args adds a second formula key.

The fix is one line: drop base.args$formula before merging when lm.args already provides its own formula, so the user-supplied formula wins cleanly.

Reproduction

library(ggformula)
data(KidsFeet, package = "mosaicData")

# Before fix: error "matched by multiple actual arguments"
# After fix: draws a quadratic fit
gf_point(length ~ width, data = KidsFeet) |>
  gf_lm(lm.args = list(formula = y ~ poly(x, 2)))

Note: the formula = y ~ poly(x, 2) direct-param syntax already works; this fixes the lm.args path to be consistent.

Test plan

• Verify lm.args = list(formula = ...) no longer errors
• Verify lm.args = list(formula = ...) produces the same fit as formula = ... directly
• Verify default behavior (lm.args = list()) is unchanged

🤖 Generated with Claude Code

[rpruim]

Thanks.

I'm traveling the next two weeks, and I may not get a chance to look at this until I return.