get rid of one sage_eval in complex numbers

[fchapoton]

this is parsing in a simple way the string input for complex number constructor, avoiding usage of sage_eval there

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

[github-actions]

Documentation preview for this PR (built with commit 191dd38; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[user202729]

Depends on how much external code uses the string constructor of ℂ, this may break badly.

Here are some examples

sage: CC("e")
2.71828182845905
sage: CC("e**j")
0.540302305868140 + 0.841470984807897*I
sage: CC("3+I")
3.00000000000000 + 1.00000000000000*I
sage: CC("I**2")
-1.00000000000000
sage: CC("12 34")
1234.00000000000

Also, is it faster than sage_eval (or just do a eval?)

[fchapoton]

This parser is not supposed to perform computations ; it should only accept inputs that are already just one complex number in full explicit form.

[fchapoton]

after

sage: %timeit CC('13.67788e26+1.20958686e-14*I')
2.83 µs ± 7.93 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

versus before

sage: %timeit CC('13.67788e26+1.20958686e-14*I')
27.3 µs ± 151 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

[mkoeppe]

LGTM. Thanks for working on this!