Expanded on docstrings related to sampling in acquisition functions (#2010)

Summary:
## Motivation

I did this mainly for my own understanding.

### Have you read the [Contributing Guidelines on pull requests](https://github.com/pytorch/botorch/blob/main/CONTRIBUTING.md#pull-requests)?

Yes

Pull Request resolved: #2010

Test Plan: Human review

Reviewed By: Balandat

Differential Revision: D49120984

Pulled By: esantorella

fbshipit-source-id: d9ad0df21e34d9d46c20eefeac720d00a944b823

Author: esantorella

botorch/acquisition/acquisition.py

@@ -124,7 +124,8 @@ def __init__(self, sampler: Optional[MCSampler] = None) -> None:
 
         Args:
             sampler: The sampler used to draw base samples for MC-based acquisition
-                functions. If `None`, a sampler is generated using `get_sampler`.
+                functions. If `None`, a sampler is generated on the fly within
+                the `get_posterior_samples` method using `get_sampler`.
         """
         self.sampler = sampler
 

botorch/acquisition/monte_carlo.py

@@ -70,7 +70,9 @@ def __init__(
         Args:
             model: A fitted model.
             sampler: The sampler used to draw base samples. If not given,
-                a sampler is generated using `get_sampler`.
+                a sampler is generated on the fly within the
+                `get_posterior_samples` method using
+                `botorch.sampling.get_sampler`.
                 NOTE: For posteriors that do not support base samples,
                 a sampler compatible with intended use case must be provided.
                 See `ForkedRNGSampler` and `StochasticSampler` as examples.
@@ -189,8 +191,10 @@ def __init__(
 
         Args:
             model: A fitted model.
-            sampler: The sampler used to draw base samples. If not given,
-                a sampler is generated using `get_sampler`.
+            sampler: The sampler used to draw base samples. If not given, a
+                sampler is generated on the fly within the
+                `get_posterior_samples` method using
+                `botorch.sampling.get_sampler`.
                 NOTE: For posteriors that do not support base samples,
                 a sampler compatible with intended use case must be provided.
                 See `ForkedRNGSampler` and `StochasticSampler` as examples.
@@ -557,7 +561,7 @@ def compute_best_f(self, obj: Tensor) -> Tensor:
         return val.view(view_shape).to(obj)
 
     def _sample_forward(self, obj: Tensor) -> Tensor:
-        r"""Evaluate qNoisyExpectedImprovement per sample on the candidate set `X`.
+        """Evaluate qNoisyExpectedImprovement per objective value in `obj`.
 
         Args:
             obj: A `sample_shape x batch_shape x q`-dim Tensor of MC objective values.

botorch/optim/optimize.py

@@ -217,6 +217,9 @@ def _optimize_acqf_sequential_q(
 ) -> Tuple[Tensor, Tensor]:
     """
     Helper function for `optimize_acqf` when sequential=True and q > 1.
+
+    For each of `q` times, generate a single candidate greedily, then add it to
+    the list of pending points.
     """
     _validate_sequential_inputs(opt_inputs)
     # When using sequential optimization, we allocate the total timeout