Skip UTF-8 BOM before parsing

[bcmeireles]

skip initial UTF-8 BOM before parsing text input and grammar files. Fixes #407

[MegaIng]

IMO this isn't larks job to do:

• utf-8 BOM is very windows specific and generally not recommend - to the point where one can say input is malformed if it contains a BOM
• It's part of the file encoding, which is not the abstraction layer lark operates on. Open files with the utf-8-sig encoding and the bom will be removed by the encoding layer built into python (See also: https://discuss.python.org/t/utf-8-bom-not-being-consumed-when-opening-file/74870)
• We especially shouldn't do this for random bytes input where we don't even know if it's intended to be text.
• (By using slicing we are copying the data which is something we are explicitly trying to avoid. This also falsifies indices. This could be fixed, but I still don't think we should for the reasons above)