Fix algo option type conversion being silently skipped with postponed annotations

[janosg]

Problem

Algorithm.__post_init__ converts and validates algo option values by looking up each dataclass field's type in TYPE_CONVERTERS, which is keyed by type objects (PositiveInt, int, ...). In modules with from __future__ import annotations — which the how-to-document-optimizers guide requires so that autodoc renders type aliases readably — field.type is the annotation string ("PositiveInt"), so the lookup never matched and coercion/validation was silently skipped.

This already affected every optimizer module that uses the import on main (scipy, gfo, nevergrad, pyswarms, pygad, bayesian, pygmo, tranquilo, iminuit) and surfaced as a test failure in #690 when ipopt.py gained the import: test_ipopt_algo_options passes 5.5 for an integer-typed Ipopt option and relies on coercion to int.

Fix

• TYPE_CONVERTERS_BY_NAME in type_conversion.py: a string-keyed companion table with the same converters.
• Algorithm.__post_init__ branches on whether field.type is a string and uses the matching table, so coercion works identically with and without the future import.

Tests

• test_algorithm_future_annotations.py (new): a dummy algorithm defined under postponed annotations, mirroring the existing type conversion tests in test_algorithm.py, plus a guard test asserting the field types really are annotation strings.
• test_algorithm.py: a mirror guard asserting the existing dummy's field types are type objects (so the two modules keep covering both code paths), a float→int coercion test, and test_int_options_are_coerced_for_all_algorithms, which passes floats for all int-typed options of every algorithm in ALL_ALGORITHMS and asserts they come back as ints.

Without the fix, 97 of the new tests fail (the postponed-annotations module and the registry-wide test for ~90 algorithms); with it, everything passes.

#690 is rebased on top of this branch.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/optimagic/optimization/algorithm.py	95.04% <100.00%> (+0.07%)	⬆️
src/optimagic/type_conversion.py	100.00% <100.00%> (ø)

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.