feat(view): reserve capacity for packed repeated fields (#171)

[yanmo42]

This PR implements capacity reservation for packed repeated scalar view fields as suggested in #171.

Performance

Using the reproduction provided in #171, this change reduces decoding time by approximately 30.4% (7.94s -> 5.53s) on my environment by eliminating redundant reallocations during view creation.

Implementation

• Added RepeatedView::reserve to the public (but hidden) view API.
• Updated codegen to emit .reserve() calls before decoding packed fields.
• Fixed-size types (fixed32, fixed64, float, double) reserve the exact required capacity.
• Varint types reserve capacity equal to the byte length of the payload (a safe upper bound).

Validation

• Unit tests for RepeatedView::reserve passed.
• Codegen test test_view_packed_scalar_reserves_capacity passed.
• All core buffa and buffa-codegen unit tests passed.
• Clippy and fmt passed for buffa and buffa-codegen.

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[yanmo42]

I have read the CLA Document and I hereby sign the CLA

[yanmo42]

submitted prematurely, will re-evaluate