Keep logger names in hash set

Previously, the list of logger names that have been created in a given
context with Log4J was being kept in a `Vector`. As a result, subsequent
calls to `logger` on a `Log4JLoggerContext` with the same logger name
would introduce duplicates in the list. Note that Log4J's logger
configuration would not allow duplicates, and only one configuration
object for a given logger name will exist. The duplicates in the list of
logger names are therefore unnecessary.

With this patch, the logger names will be kept in a `HashSet`.

Since the logger names are only used to remove the logger configurations
after the logger context is closed, and the logger configurations
contain no duplicates, removing duplicate logger names should not cause
any issue.

internal/util-logging/src/main/scala/sbt/util/LoggerContext.scala

@@ -46,7 +46,7 @@ object LoggerContext {
       case a: AbstractConfiguration => a
       case _                        => throw new IllegalStateException("")
     }
-    val loggers = new java.util.Vector[String]
+    val loggers = new java.util.HashSet[String]
     private[this] val closed = new AtomicBoolean(false)
     override def logger(
         name: String,