Merge pull request #129 from microsoft/connor4312/token-limit-element

feat: add a `TokenLimit` element

Adopt LanguageModelToolCallPart arg reordering (#125)

Fix countMessageTokens (#124)

* Fix countMessageTokens

* Version bump

Fix `countMessageTokens` (#123)

* Fix

* Version bump

Adopt finalized form of tools api, content2 -> content (#122)

* Adopt finalized form of tools api, content2 -> content

* Version bump

Export ToolResult (#116)

Merge pull request #115 from microsoft/connor4312/pass-wrapper-priority

feat: add a passPriority attribute for logical wrapper elements

fix: containers without priority set should have max priority (#109)

* fix: containers without priority set should have max priority

Closes microsoft/vscode-copilot#9289

* run validation on pr

* 0.3.0-alpha.6

feat: add Expandable elements (#105)

Expandable can be used like so, providing a callback
which returns a string or a promise to a string:

```tsx
<Expandable value={async sizing => {
  let data = 'hi';
  while (true) {
    const more = getMoreUsefulData();
    if (await sizing.countTokens(data + more) > sizing.tokenBudget) { break }
    data += more;
  }
  }
  return data;
}} />
```

After the prompt is rendered, the renderer sums up the tokens used by
all messages. If there is unused budget, then any `<Expandable />`
elements' values are called again with their `PromptSizing` is increased
by the token excess.

If there are multiple `<Expandable />` elements, then they're re-called
in the order in which they were initially rendered. Because they're
designed to fill up any remaining space, it usually makes sense to have
at most one `<Expandable />` element per prompt.

feat: enhance HTML tracer to visualize prune order (#103)

* wip

* wip

* finish up

* rm gitignored