fix: handle empty power history in CPU.total_power() (#832)

[agodianel]

Description

This PR fixes a ZeroDivisionError crash that occurs in the total_power method when the system fails to collect any power samples (resulting in an empty power history).

The previous code checked whether the power history length was 0 immediately after an append call, meaning the condition was a dead code path and never caught the empty list. This has been updated to check for an empty list of watts immediately before the division step, safely returning 0 W instead of crashing.

Related Issue

Fixes #832

Motivation and Context

Prevents CodeCarbon from crashing entirely when hardware sensors (like Intel RAPL or Power Gadget) occasionally fail to return valid readings during a sampling interval.

How Has This Been Tested?

• Added a regression test that intentionally feeds 0W measurements and clears the power history, asserting that the method handles it gracefully without a ZeroDivisionError.
• Ran the full test suite to ensure no existing functionality was broken (all CPU and core tests pass successfully).

Types of changes

What types of changes does your code introduce? Put an x in all the boxes that apply:

• Bug fix (non-breaking change which fixes an issue