Introduce Warm-up Phase for Dudect Measurement

[HeatCrab]

A warm-up step was added to the measurement function to discard the first batch of data, inspired by a privately implemented earlier version that passed tests for queue-related functions. This follows the approach described in the Dudect paper, which skips initial measurements to improve timing stability., potentially mitigating factors like cache misses or allocation delays. Unlike the author's GitHub implementation, which dynamically checks the percentiles array state for each call, a static boolean is used here to mark the first execution within a single test run, as this suits the fixed iteration structure of the testing loop. Tests using queue insertion and removal functions showed no notable difference in constant-time behavior with or without this step after memory leaks were fixed, but it is retained for potential improvements in t-test precision under different test conditions.

Change-Id: Ieec138264fdd8d087142cdf3c3e9f961f521e80b

[jserv]

The commit messages were misleading:

• inspired by a privately implemented earlier version that passed tests => What's that?
• Dudect paper => Use the fully-qualified form if you want to refer to a paper.
• the author's GitHub implementation => You should put the name of the author.
• potential improvements in t-test precision under different test conditions. => How come?

Clarify them before reviewing.