Fix logging nuisances that can arise when importing flaml

Previously, FLAML would automatically set its own loglevel when
imported, overriding any other logging settings that may previously have
been set by users of the library. FLAML will now check whether a
loglevel has been explicitly set before setting its own.

In addition, the `flaml.automl` warning is now emitted using `warnings`
rather than `logging.warning`. This falls more in line with other FLAML
submodules (e.g. `flaml.ml`).

Taken in conjunction, these two changes make it a lot easier for
downstream users to suppress warnings emitted by FLAML upon importing
the library, either by configuring logging prior to importing `flaml` or
by using the somewhat more idiomatic
```
with warnings.catch_warnings(action="ignore"):
    import flaml
```
This change also helps to ensure that FLAML's defaults won't override
users' explicit logging settings.

flaml/__init__.py

@@ -1,4 +1,5 @@
 import logging
+import warnings
 
 try:
     from flaml.automl import AutoML, logger_formatter
@@ -12,7 +13,8 @@
 
 # Set the root logger.
 logger = logging.getLogger(__name__)
-logger.setLevel(logging.INFO)
+if logger.level == logging.NOTSET:
+    logger.setLevel(logging.INFO)
 
 if not has_automl:
-    logger.warning("flaml.automl is not available. Please install flaml[automl] to enable AutoML functionalities.")
+    warnings.warn("flaml.automl is not available. Please install flaml[automl] to enable AutoML functionalities.")