[docs] improve docs about `nthreads` parameter (#4756)

* in predict(), respect params set via `set_params()` after fit()

* extract docs changes

docs/Parameters.rst

@@ -183,6 +183,8 @@ Core Parameters
 
 -  ``num_threads`` :raw-html:`<a id="num_threads" title="Permalink to this parameter" href="#num_threads">&#x1F517;&#xFE0E;</a>`, default = ``0``, type = int, aliases: ``num_thread``, ``nthread``, ``nthreads``, ``n_jobs``
 
+   -  used only in ``train``, ``prediction`` and ``refit`` tasks or in correspondent functions of language-specific packages
+
    -  number of threads for LightGBM
 
    -  ``0`` means default number of threads in OpenMP

include/LightGBM/config.h

@@ -192,6 +192,7 @@ struct Config {
   std::string tree_learner = "serial";
 
   // alias = num_thread, nthread, nthreads, n_jobs
+  // desc = used only in ``train``, ``prediction`` and ``refit`` tasks or in correspondent functions of language-specific packages
   // desc = number of threads for LightGBM
   // desc = ``0`` means default number of threads in OpenMP
   // desc = for the best speed, set this to the number of **real CPU cores**, not the number of threads (most CPUs use `hyper-threading <https://en.wikipedia.org/wiki/Hyper-threading>`__ to generate 2 threads per CPU core)

python-package/lightgbm/sklearn.py

@@ -434,7 +434,7 @@ class LGBMModel(_LGBMModelBase):
             If RandomState object (numpy), a random integer is picked based on its state to seed the C++ code.
             If None, default seeds in C++ code are used.
         n_jobs : int, optional (default=-1)
-            Number of parallel threads.
+            Number of parallel threads to use for training (can be changed at prediction time).
         silent : bool, optional (default=True)
             Whether to print messages while running boosting.
         importance_type : str, optional (default='split')