Harmonize position of by argument

[strengejacke]

@DominiqueMakowski

One thing I always forget is that estimate_relation() and alike have by not at 2nd, but 3rd position. The second argument is data. For estimate_means() and alike, by is the second argument. I often use estimate_relation(model, "variable"), and forget to add by = , which returns an error.

Should we harmonize this? Will be a breaking change, but if we decide to do so, now would be a good opportunity, I guess. I would prefer the order model, by and then data. WDYT?

Originally posted by @strengejacke in #369 (comment)

[DominiqueMakowski]

Thanks for reminder!

I'm not a 100% convinced, but this is probably because of my habits (I typically do like:

model |>
  estimate_relation(newdata)

So I would actually keep it as is...
+ one could also argue that it's much clearer if by= is written explicitly (and it's a nice and short argument) :)

So basically it boils down to whether we want to cater to Daniel's inflexible habits ^^

[strengejacke]

Ok, let's close this. Not that crucial.