fix: expand Anthropic detection and strip whitespace-only text blocks

[kiennq]

What does this PR do?

When using Anthropic models through non-standard provider configurations (e.g.Github Copilot), the existing check model.api.npm === "@ai-sdk/anthropic" fails to detect them as Anthropic. This causes whitespace-only or empty text blocks to be sent to the API, which Anthropic rejects.

This PR:

• Expand Anthropic provider detection in normalizeMessages to cover all Claude/Anthropic model variants (checking providerID, api.id, model.id, and api.npm), not just @ai-sdk/anthropic
• Change empty content filtering from exact empty string (=== "") to whitespace-trimmed check (.trim() === ""), so whitespace-only text and reasoning blocks are also stripped

Should fix #2655

How did you verify your code works?

Manually test, verified I can continue with Claude model after a tool call cancel somehow insert an empty text block to request.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Based on my search, I found two potentially related PRs that address similar issues with message normalization and content filtering:

1. PR fix: strip incompatible thinking blocks when switching to Claude #8958 - "fix: strip incompatible thinking blocks when switching to Claude"

   • Related because it also deals with stripping content blocks (thinking blocks) when working with Claude/Anthropic models

2. PR fix: strip reasoning parts when switching to non-interleaved models #11572 - "fix: strip reasoning parts when switching to non-interleaved models"

   • Related because it addresses stripping reasoning parts during model switching, which overlaps with the whitespace-only text block filtering in the current PR

These PRs appear to address related concerns around message content filtering and handling Claude/Anthropic model compatibility, though they may be focused on different specific scenarios (thinking/reasoning blocks vs. whitespace-only blocks).