fix(providers): normalize mid-stream Anthropic failures to a typed error (#4)

[Hustler490]

Closes #4.

What

AnthropicProvider.stream() translated the initial POST failure into a ProviderError, but read the response body outside any handler: a connection dropped mid-stream leaked
equests.exceptions.ChunkedEncodingError to the SSE layer instead of a typed error. This mirrors the OpenAI provider fix.

• �pp/services/providers/anthropic.py: wrap the body-reading loop in ry/except requests.RequestException and re-raise ProviderUnavailableError with a safe message. The delta sequence and content_block_delta handling are unchanged.
• ests/test_providers.py: adds est_stream_mid_stream_failure_raises_typed_error (feeds one valid delta, then raises ChunkedEncodingError; asserts the already-yielded chunk is delivered and the error is a ProviderUnavailableError). Mocks only - no live network.

Acceptance criteria

• Non-streaming chat() returns a valid ProviderResponse (unchanged; pinned by existing tests).
• Streaming integrates with the SSE layer (unchanged delta behaviour).
• Anthropic SDK errors map to typed ProviderErrors - now including mid-stream transport failures.
• Unit tests use mocks; no live network calls.

[drips-wave]

@Hustler490 Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits