Data race on LoggerContext.size

[hcantunc]

LoggerContext.size has a data race. getLogger() synchronizes on the current logger tree node,
but concurrent callers from different threads hold different monitors, so incSize() → size++
is unprotected. ThreadSanitizer reports:

==================
WARNING: ThreadSanitizer: data race (pid=246682)
  Write of size 4 at 0x0006c0043a10 by thread T1908 (mutexes: write M0):
    #0  ch.qos.logback.classic.LoggerContext.incSize()         LoggerContext.java:167
    #1  ch.qos.logback.classic.LoggerContext.getLogger(String) LoggerContext.java:156
    #2  ch.qos.logback.classic.LoggerContext.getLogger(String) LoggerContext.java:53
    #3  org.slf4j.LoggerFactory.getLogger(String)              LoggerFactory.java:363
    #4  org.slf4j.LoggerFactory.getLogger(Class)               LoggerFactory.java:388
    #5  ... 

  Previous write of size 4 at 0x0006c0043a10 by thread T1915 (mutexes: write M1):
    #0  ch.qos.logback.classic.LoggerContext.incSize()         LoggerContext.java:167
    #1  ch.qos.logback.classic.LoggerContext.getLogger(String) LoggerContext.java:156
    #2  ch.qos.logback.classic.LoggerContext.getLogger(String) LoggerContext.java:53
    #3  org.slf4j.LoggerFactory.getLogger(String)              LoggerFactory.java:363
    #4  org.slf4j.LoggerFactory.getLogger(Class)               LoggerFactory.java:388
    #5  ... 
==================