Support for colon characters in PostgreSQL socket path.

Google Cloud SQL instances are now using the following format for the name
of their socket <PROJECT-ID>:<REGION>:<INSTANCE_NAME>. We do that by
allowing to escape a colon in the socket directory name by doubling it, as
in the username field. It also allows to accept any character in the socket
directory name, which is a good cleanup.

Fix #621.

pgloader.1

@@ -830,6 +830,20 @@ postgresql://unix:/tmp:54321/dbname
 .
 .IP
 The \fInetloc\fR defaults to the value of the \fBPGHOST\fR environment variable, and if it is unset, to either the default \fBunix\fR socket path when running on a Unix system, and \fBlocalhost\fR otherwise\.
+.
+.IP
+Socket path containing colons are supported by doubling the colons within the path, as in the following example:
+.
+.IP "" 4
+.
+.nf
+
+postgresql://unix:/tmp/project::region::instance:5432/dbname
+.
+.fi
+.
+.IP "" 0
+
 .
 .IP "\(bu" 4
 \fIdbname\fR

pgloader.1.md

@@ -704,6 +704,11 @@ Where:
     variable, and if it is unset, to either the default `unix` socket path
     when running on a Unix system, and `localhost` otherwise.
 
+    Socket path containing colons are supported by doubling the colons
+    within the path, as in the following example:
+    
+        postgresql://unix:/tmp/project::region::instance:5432/dbname
+
   - *dbname*
 
 	Should be a proper identifier (letter followed by a mix of letters,

src/parsers/command-db-uri.lisp

@@ -60,7 +60,8 @@
   (or (member char #.(quote (coerce "/.-_" 'list)))
       (alphanumericp char)))
 
-(defrule socket-directory (and "unix:" (* (socket-directory-character-p character)))
+(defrule socket-directory (and "unix:"
+                               (* (or (not ":") doubled-colon)))
   (:destructure (unix socket-directory)
 		(declare (ignore unix))
     (list :unix (when socket-directory (text socket-directory)))))