Fix getTokensFromNode template scanning issue

[Copilot]

The getTokensFromNode function in internal/ls/signaturehelp.go was incorrectly tokenizing template expressions, producing NoSubstitutionTemplateLiteral tokens that spanned across template boundaries instead of recognizing proper template structure.

Issue

When scanning a template expression like:

tag`Hello ${value} world ${value}!`

The function would incorrectly produce a NoSubstitutionTemplateLiteral token from the closing backtick to the next token (e.g., ";" instead of separate "`" and ";" tokens). This happened because the scanner wasn't aware of template context when starting from a specific position within a template expression.

Root Cause

The scanner would interpret closing backticks as the beginning of new NoSubstitutionTemplateLiteral tokens rather than TemplateTail tokens, because it lacked the proper template context when initialized at arbitrary positions within template expressions.

Solution

Added special handling for TemplateExpression nodes in getTokensFromNode:

• Recognizes when processing a template expression AST node
• Uses the AST structure to generate proper tokens for template parts
• Returns correct sequence: TemplateHead → expressions → TemplateMiddle → expressions → TemplateTail
• Recursively processes expression parts within template spans

Verification

The fix correctly tokenizes the example above as:

1. TemplateHead: "Hello ${"
2. Expression: value
3. TemplateMiddle: "} world ${"
4. Expression: value
5. TemplateTail: "}!"`

All existing tests pass and signature help for tagged templates now works correctly.

Fixes #1554.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.