Simplify (and slightly speed up) coefb/coefa

[martinholters]

Noticed this the other day. Main benefit is readability. Only in cases where trailing/leading zeros had to be re-appended/prepended, one allocation is eliminated. Otherwise, performance is almost unchanged.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.96%. Comparing base (16f1ec0) to head (6e55a5e).
Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #607   +/-   ##
=======================================
  Coverage   97.96%   97.96%           
=======================================
  Files          19       19           
  Lines        3248     3248           
=======================================
  Hits         3182     3182           
  Misses         66       66           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[wheeheee]

Somewhat forgot how the indexing into LaurentPolynomial works, but are there a lot of branches to getindex the leading/trailing zeros? Maybe that's bad for branch prediction or something. Anyhow, saving one allocation is probably better.

[martinholters]

Well, yes, the getindex branches on whether the requested index is inside the stored coefficients, but Polynomials.coeffs also calls getindex internally, so not much difference there.