feat(core)!: Remove enableTruncation flag and gen_ai text truncation

Removes the `enableTruncation` integration option and the gen_ai message
text-truncation machinery. Now that gen_ai spans always ride the v2 span
path (far larger size limits than legacy transactions), the byte-limit
cropping and "keep only the last message" reduction is dead weight.

Inline media stripping is kept and is now unconditional: previously media
was only stripped on the truncation path, so with truncation off inline
base64 could leak into span attributes for openai/anthropic/google-genai/
vercel-ai/workers-ai. `getGenAiMessagesJsonString` now always strips media
before serializing, keeping all messages.

Media-stripping logic is consolidated into `mediaStripping.ts`;
`messageTruncation.ts` is deleted. Truncation-specific test scenarios and
instruments are removed; image-stripping tests are re-pointed at the
default recording instruments, and full span-streaming coverage is kept
via non-truncation streaming tests.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: nicohrubec

dev-packages/node-integration-tests/suites/tracing/anthropic/instrument-no-truncation.mjs

@@ -36,7 +36,6 @@ async function run() {
     });
 
     // Send the image showing the number 3
-    // Put the image in the last message so it doesn't get dropped
     await client.messages.create({
       model: 'claude-3-haiku-20240307',
       max_tokens: 1024,