Ensure parser always returns `erl_anno:location()` in errors

[gomoripeti]

This way the return type matches error_info()

This is an alternative fix for https://github.com/WhatsApp/erlfmt/pull/324

[michalmuskala]

Hey, sorry for taking so long to reply on this

I think it would be better to do the reverse - to make sure, if possible, we always return the richer types on errors. If anything the errors are more informative with potential start & end location of the error, and not just start one.

[gomoripeti]

thanks for the feedback

some pros for the simple location

• consistency with erlfmt_scan which also returns a location only https://github.com/WhatsApp/erlfmt/blob/main/src/erlfmt_scan.erl#L73 (I don't know if an end_location makes sense at all in case of tokenizing)
• consistency with OTP libs like https://www.erlang.org/doc/man/compile.html#error_information, https://www.erlang.org/doc/man/yecc.html#data-types etc (ofc erlfmt forked the scanner/parser to get rid of the limitations of OTP)
• consistent response in error_info() of erlfmt:read_nodes[_string])

Which one would you prefer?

• just allow both location and anno in the type of erlfmt:error_info() (as in #324)?
• modify erlfmt_scan to also return an anno (end_location is a mandatory field of erlfmt_scan:anno so then end_location = start_location?). also modify the formatting code (which I am absolutely note familiar with)
• modify erlfmt module only to unify the error_infos coming from different components