Python: Fix non-ascii chars in span attributes

[TaoChenOSU]

Motivation and Context

Closes #3495

Description

If something contains non-ascii strings and when it's serialized to Json, the non-ascii chars will get escaped to unicodes that are not human readable.

This PR makes escaping non-ascii chars not enforced.

Before:

After:

Note that depending on the encoding used by monitor backend or console, texts may still be escaped.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework
_tools.py	755	84	88%	175–176, 306, 308, 326–328, 335, 353, 367, 379, 384, 386, 393, 430, 453–455, 504–506, 569, 591, 615–641, 676, 684, 925, 1187, 1244, 1248, 1327–1331, 1349, 1351–1352, 1464, 1468, 1518, 1520, 1536, 1538, 1602, 1629, 1686, 1754, 1933–1934, 1961, 1969, 1982, 1992–1993, 2028, 2084, 2116
observability.py	618	84	86%	335, 337–339, 342–344, 349–350, 356–357, 363–364, 371, 373–375, 378–380, 385–386, 392–393, 399–400, 407, 676, 679, 687–688, 691–694, 696, 699–701, 704–705, 733, 735, 746–748, 750–753, 757, 765, 866, 868, 1017, 1019, 1023–1028, 1030, 1033–1037, 1039, 1151–1152, 1154, 1211–1212, 1347, 1401–1402, 1518–1520, 1579, 1749, 1903, 1905
TOTAL	21258	3276	84%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
4106	225 💤	0 ❌	0 🔥	1m 14s ⏱️

[Copilot]

Pull request overview

This PR fixes an issue where non-ASCII characters (Japanese, Chinese, Arabic, Korean, etc.) in OpenTelemetry span attributes were being escaped to Unicode sequences (e.g., \u4f01\u696d), making them unreadable in telemetry backends. The fix adds ensure_ascii=False to json.dumps() calls when serializing messages, system instructions, and tool definitions for telemetry.

Changes:

• Added ensure_ascii=False parameter to JSON serialization calls in observability code for messages, system instructions, and tool definitions
• Added ensure_ascii=False parameter to JSON serialization of tool arguments
• Added comprehensive test coverage for non-ASCII character preservation across messages, system instructions, tool arguments, and tool results

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
python/packages/core/agent_framework/observability.py	Added ensure_ascii=False to json.dumps calls for tool definitions (line 1560), input/output messages (line 1643), and system instructions (line 1649)
python/packages/core/agent_framework/_tools.py	Added ensure_ascii=False to json.dumps call for tool arguments when not using Pydantic model (line 480)
python/packages/core/tests/core/test_observability.py	Added 4 comprehensive tests covering non-ASCII character preservation in messages, system instructions, tool arguments, and tool results