When lexing a source buffer with a non-zero offset, e.g. due to a
skipped interpreter line, lexical tokens reported a wrong offset
to the compiler, causing error locations and source context lines
to be incorrectly shifted.

-- Testcase --
#!/usr/bin/env ucode
{%
	die("Error");
%}
-- End --

-- Expect stderr --
Error
In line 3, byte 12:

 `    die("Error");`
  Near here -----^


-- End --