Add warning about precision in interfaces documentation (#8124)

**Context:**
A user query on our forum ([Thread
#8885](https://discuss.pennylane.ai/t/pytorch-default-data-type/8885))
revealed that PennyLane silently promotes all float32 inputs to float64
during computation.

**Description of the Change:**

This change adds a warning to the [Interfaces
](https://docs.pennylane.ai/en/stable/introduction/interfaces.html)documentation
regarding the behaviour (everything is promoted to 64-bit outputs).

**Benefits:**

**Possible Drawbacks:**

**Related GitHub Issues:**

---------

Co-authored-by: Catalina Albornoz <[email protected]>
Co-authored-by: Isaac De Vlugt <[email protected]>

Author: 3 people

doc/introduction/interfaces.rst

@@ -118,6 +118,11 @@ PennyLane also provides higher-level classes for converting QNodes into ``torch.
 
     pennylane.qnn.TorchLayer
 
+.. warning::
+
+    PennyLane's QNodes currently promote all ``torch.float32`` (single-precision) inputs to ``torch.float64`` 
+    (double-precision) during execution. This may result in higher memory usage than expected.
+
 
 .. note::
 

doc/releases/changelog-dev.md

@@ -1034,6 +1034,10 @@
   Trimmed the outdated part of discussion regarding different choices of `alpha`.
   [(#8100)](https://github.com/PennyLaneAI/pennylane/pull/8100)
 
+* A warning was added to the :doc:`interfaces documentation </introduction/interfaces>` under the Pytorch section saying that all Pytorch floating-point inputs are promoted 
+  to `torch.float64`.
+  [(#8124)](https://github.com/PennyLaneAI/pennylane/pull/8124)
+
 * The :doc:`Dynamic Quantum Circuits </introduction/dynamic_quantum_circuits>` page has been updated to include the latest device-dependent mid-circuit measurement method defaults.
   [(#8149)](https://github.com/PennyLaneAI/pennylane/pull/8149)
 
@@ -1149,6 +1153,7 @@ Pablo Antonio Moreno Casares
 Erick Ochoa,
 Mudit Pandey,
 Andrija Paurevic,
+Justin Pickering,
 Alex Preciado,
 Shuli Shu,
 Jay Soni,