fix osm_line2poly

@richardbeare I've gone through most of your code and vignette with this commit. Note the following in relation to your vignette:

1. osmdata::opq() now accepts explicit timeout (and memsize) arguments, but these actually aren't needed here. It should just work as is, and does for me, so no need to distract readers of vignette with that stuff
2. I've removed most of the data you previously had (all of the highways), because this made the plot too large to include in an R package. We really only need the coastal data to illustrate the functionality, and I actually think this really aids the focus of the whole thing.
3. This should actually make it small enough to include as data within the package, thereby enabling the vignette to actually be run without the graphs.

Now to the issue here which I'd love your help with:

tidy line2poly.R

I'm guessing you don't work with a code linter? Most of the changes I've made are just standard lintr suggestions, with a few extra personal things (more whitespace; vertical alignment of curly braces). I also put all separate functions out into the main function space. However, in it's current form this function no longer works. Can you just have a fish around to ensure that it all works. I think one problem is with the bbox interpretation - you often presume it has a fixed, matrix form with names rows and columns, yet there are no checks for this. Could you just write a small function to force any differently-formed bbox args into this form? (Or some equivalent way of dealing with this). Thanks in advance, and thanks for all the help here - very much appreciated!

Start with line2poly.R and review the linter-related changes described in the issue, especially the bbox assumptions. Check how differently formed bbox arguments are handled and verify that osm_line2poly works again. Done means the function runs correctly with the relevant bbox forms without the noted tidy-up breaking it.