Detect malformed * notation in collation rules

We have the following known issue in ICU 4.6: A collation rule using the new star notation “<*” in which the list of characters following “<*” includes two successive APOSTROPHE characters ( ' ' ) is malformed, but may not be detected and will produce erroneous results.

So we need to detect this in ucol_openRules and the like. We may not need to check for this in genrb if LDML2ICUConverter is changed to never generate this.