Fix performace issue

[372046933]

GCC default optimization level is 0, i.e. no optimization. Production environment usually has optimization level at least 2.
When -O3 is used, cqueue is three times slower than std::deque. The root cause is modulo operator. Every push/pop needs to modulo reserved_.
Since reserved_ is always power of 2. Modulo can be replaced by bit operation.
Before the PR

g++ -std=c++20 -g -03 -o deque-prof deque-prof.cpp
g++ -std=c++20 -g -03 -o cqueue-prof cqueue-prof.cpp
./deque-prof Elapsed time in microseconds : 63234 us
./cqueue-prof
Elapsed time in microseconds : 224606 us

After:

./deque-prof Elapsed time in microseconds : 63064 us
./cqueue-prof
Elapsed time in microseconds : 33249 us

[torrentg]

Impressive improvement !
I slightly modified the suggested change to consider the case mReserved != power of 2.
Thank you for your contribution.

[372046933]

Thank you for your review. Does shrink_to_fit break the power of 2 guarantee? Since growth is always power of 2

[torrentg]

You are welcome. Yes, shrink_to_fit and the constructor with fixed capacity breaks the power of two garantee. This was reported by some tests that failed. I suspect that usual usage case don't use this features, this is because I set a [likely] in the power of two case.