[uri] Apply URI decoding for all parsed URIs

The various early-exit paths in parse_uri() accidentally bypass the
URI field decoding.  The result is that opaque or relative URIs do not
undergo URI field decoding, resulting in double-encoding when the URIs
are subsequently used.  For example:

  #!ipxe
  set mac ${macstring}
  imgfetch /boot/by-mac/${mac:uristring}

would result in an HTTP GET such as

  GET /boot/by-mac/00%253A0c%253A29%253Ac5%253A39%253Aa1 HTTP/1.1

rather than the expected

  GET /boot/by-mac/00%3A0c%3A29%3Ac5%3A39%3Aa1 HTTP/1.1

Fix by ensuring that URI decoding is always applied regardless of the
URI format.

Reported-by: Andrew Widdersheim <awiddersheim@inetu.net>
Signed-off-by: Michael Brown <mcb30@ipxe.org>

src/core/uri.c

 		uri->port = tmp;
 	}
 
+ done:
 	/* Decode fields in-place */
 	for ( field = 0 ; field < URI_FIELDS ; field++ )
 		uri_decode_inplace ( uri, field );
 
- done:
 	DBGC ( uri, "URI parsed \"%s\" to", uri_string );
 	uri_dump ( uri );
 	DBGC ( uri, "\n" );

src/tests/uri_test.c

 	{ .scheme = "mailto", .opaque = "ipxe-devel@lists.ipxe.org" }
 };
 
+/** Basic path-only URI */
+static struct uri_test uri_path = {
+	"/var/lib/tftpboot/pxelinux.0",
+	{ .path = "/var/lib/tftpboot/pxelinux.0" },
+};
+
+/** Path-only URI with escaped characters */
+static struct uri_test uri_path_escaped = {
+	"/hello%20world%3F",
+	{ .path = "/hello world?" },
+};
+
 /** HTTP URI with all the trimmings */
 static struct uri_test uri_http_all = {
 	"http://anon:password@example.com:3001/~foo/cgi-bin/foo.pl?a=b&c=d#bit",
 	uri_parse_format_dup_ok ( &uri_empty );
 	uri_parse_format_dup_ok ( &uri_boot_ipxe_org );
 	uri_parse_format_dup_ok ( &uri_mailto );
+	uri_parse_format_dup_ok ( &uri_path );
+	uri_parse_format_dup_ok ( &uri_path_escaped );
 	uri_parse_format_dup_ok ( &uri_http_all );
 	uri_parse_format_dup_ok ( &uri_http_escaped );
 	uri_parse_ok ( &uri_http_escaped_improper ); /* Parse only */