Fix/3860 nested formatter precision

[way4sahil]

Fixes #3860

Dynamic width and precision specifications inside nested_formatter now
resolve argument IDs against the enclosing format context.

This adds:

• Context forwarding for nested formatters.
• Dynamic width handling.
• Named and explicit argument support.
• Tests for dynamic width, precision, padding, and static formatting.
• A ChangeLog entry.

[fallenmi]

Requesting changes at exact head ee0b93c72d2ecd933d24997304bf6e6de2488a0a.

The dynamic-precision path itself works when the formatter uses the new nested(value, ctx) overload: my exact-base oracle (21d4cc15c8a2a3536aab5623258ba3b3dae573ce) reports argument not found, while this head produces (1.00, 2.00). The focused head test also passes (format-test --gtest_filter=format_test.nested_formatter, 1/1, 9.5 s including configure/build).

There are two attributable blockers on this head:

1. The repository's g++-13 c++23 Release Shared job fails in format.h:4178: the new local alias using context = ... shadows the existing fmt::context, and the project promotes -Wshadow to an error. Please rename that alias (for example, nested_context_type).
2. The clang-format job fails around the same specialization. I reproduced this with Apple clang-format 21 and the repository config: the exact base produces no diff, while the exact head rewrites lines 4171-4181. Please apply clang-format 21 to the touched code.

Please also drop unrelated commit d0d9adaf6ee72213ad2f9925b7056d5eb2890467 from this branch. Its C++20 module documentation change is already being reviewed separately in #4891 and is unrelated to the nested_formatter fix.

Disclosure: this review was prepared with Codex assistance; I verified the diff, exact SHAs, test outputs, and current CI logs before submitting it.