Add support for right-sided formulas as arguments `clusters` and `weights`

[rossellhayes]

This PR addresses #380 by adding support for right-sided formulas as arguments clusters and weights in model functions. With this change, clusters and weights can more easily be set by other functions calling estimatr models, solving the motivating problem in #380:

purrr::map(
  list(~ am, ~ cyl),
  ~ estimatr::lm_robust(mpg ~ hp, data = mtcars, clusters = .)
)
#> [[1]]
#>                Estimate  Std. Error   t value   Pr(>|t|)   CI Lower    CI Upper
#> (Intercept) 30.09886054 2.450272545  12.28388 0.05171030 -1.0340203 61.23174140
#> hp          -0.06822828 0.004484306 -15.21490 0.03879625 -0.1215026 -0.01495395
#>                   DF
#> (Intercept) 1.000011
#> hp          1.029391
#> 
#> [[2]]
#>                Estimate Std. Error   t value   Pr(>|t|)   CI Lower  CI Upper
#> (Intercept) 30.09886054 5.04648516  5.964322 0.05856662 -3.7089449 63.906666
#> hp          -0.06822828 0.02652462 -2.572262 0.17539476 -0.2460305  0.109574
#>                   DF
#> (Intercept) 1.394247
#> hp          1.393706

^{Created on 2021-04-09 by the reprex package (v1.0.0)}

[nfultz]

See also my comment on the ticket - if the workaround is sufficient, no need to merge - https://github.com/DeclareDesign/estimatr/issues/380#issuecomment-817151396

[rossellhayes]

Agreed, using quos was able to handle my use case, so I think that should be sufficient.