Perf: Optimize string concatenation in parse_uncertainty

[mohasarc]

This pull request optimizes string construction in parse_uncertainty by replacing + concatenation with .join().

The Issue: In Python, strings are immutable. Using + to combine multiple string segments forces the interpreter to create a new string object and copy the content for every concatenation step. This results in inefficient memory usage and unnecessary overhead, particularly when constructing strings from multiple parts.

The Solution: I refactored the string construction logic to use ''.join(). This approach ensures linear time complexity O(n) because the total size of the final string is calculated once, and memory is allocated efficiently.

This is in line with standard python performance recommendation: https://peps.python.org/pep-0008/#programming-recommendations

[pkienzle]

On my machine the new version is slower. I also find it a little less readable.

value, unc = "3.1415913", "15"

%timeit zeros = len(value.split('.')[1]) - len(unc); full_unc = "0." + ("0"*zeros) + unc
132 ns ± 1.89 ns per loop (mean ± std. dev. of 7 runs, 10,000,000 loops each)

%timeit zeros = len(value.split('.')[1]) - len(unc); full_unc = "".join(["0.", "0" * zeros, unc])
148 ns ± 0.603 ns per loop (mean ± std. dev. of 7 runs, 10,000,000 loops each)

If you want to update the PR using f-strings that would be okay.

%timeit zeros = len(value.split('.')[1]) - len(unc); full_unc = f"0.{'0' * zeros}{unc}"
115 ns ± 0.0236 ns per loop (mean ± std. dev. of 7 runs, 10,000,000 loops each)

At 15 ns improvement per parsed number, I don't expect this to change startup times significantly, but every little bit helps.

[edit: changed "submit a new PR" to "update the PR"]
[edit: revised timing impact statement]