Avoid doing unneeded logger work in Replication

Most of the replication log statements do some work including memory
allocations which are then thrown away if the log level it too high, so always
check the log level first. A few places don't actually benefit from this, but
it's easier to consistently check the log level every time.

CHANGELOG.md

@@ -12,7 +12,8 @@
 * Passing a double as argument for a Query on Decimal128 did not work ([#7386](https://github.com/realm/realm-core/issues/7386), since v14.0.0)
 * Opening file with file format 23 in read-only mode will crash ([#7388](https://github.com/realm/realm-core/issues/7388), since v14.0.0)
 * Querying a dictionary over a link would sometimes result in out-of-bounds memory reads ([PR #7382](https://github.com/realm/realm-core/pull/7382), since v14.0.0).
-* Restore the pre-14.0.0 behavior of missing keys in dictionaries in queries  ([PR #7391](https://github.com/realm/realm-core/pull/7391))
+* Restore the pre-14.0.0 behavior of missing keys in dictionaries in queries ([PR #7391](https://github.com/realm/realm-core/pull/7391))
+* Fix a ~10% performance regression for bulk insertion when using a log level which does not include debug/trace ([PR #7400](https://github.com/realm/realm-core/pull/7400), since v14.0.0)
 
 ### Breaking changes
 * None.