chore: eval extra's google-cloud-aiplatform bound has no upper cap, unlike all/gcp/test

🔴 Required Information

Describe the Bug:
Not a crash report — a version-constraint consistency question. pyproject.toml's eval optional-dependency extra constrains google-cloud-aiplatform with no upper bound, while three sibling extras that also depend on google-cloud-aiplatform all cap it at <2. Quoted verbatim, at origin/main c7ffcfa8:

• optional-dependencies.all (pyproject.toml:82): "google-cloud-aiplatform[agent-engines,evaluation]>=1.148.1,<2"
• optional-dependencies.eval (pyproject.toml:195): "google-cloud-aiplatform[evaluation]>=1.148" — no upper bound
• optional-dependencies.gcp (pyproject.toml:225): "google-cloud-aiplatform[agent-engines]>=1.148.1,<2"
• optional-dependencies.test (pyproject.toml:290): "google-cloud-aiplatform[agent-engines,evaluation]>=1.148.1,<2"

Three of the four extras that pull in google-cloud-aiplatform cap it below the 2.x major line; eval — the one most users installing for evaluation support (pip install google-adk[eval]) actually reach — does not.

Steps to Reproduce:

1. Install google-adk[eval] in a fresh environment with no version overrides (e.g. pip install -e ".[eval]" against a checkout of this repo, or pip install "google-adk[eval]" once published).
2. Check the resolved google-cloud-aiplatform version (pip show google-cloud-aiplatform).
3. Compare against what pip install "google-adk[all]", [gcp], or [test] would resolve for the same package.

Expected Behavior:
All four extras that depend on google-cloud-aiplatform apply the same version bound, or the difference is intentional and documented.

Observed Behavior:
pip install google-adk[eval] resolves google-cloud-aiplatform on the 2.x line (observed: 2.1.0), while [all], [gcp], and [test] all resolve to the last version under 2 (observed: 1.165.1) for the identical base requirement. No error or crash is being reported here — this issue is scoped to the constraint inconsistency itself; whether the eval extra's own missing upper bound is intentional (e.g. evaluation-specific APIs needing 2.x) or an oversight from when the bound was added to the other three extras is what's being asked.

Environment Details:

• ADK Library Version (pip show google-adk): unreleased, google/adk-python origin/main c7ffcfa8 (__version__ reports 2.8.0)
• Desktop OS: Windows 11
• Python Version (python -V): 3.13.5

Model Information:

• Are you using LiteLLM: N/A
• Which model is being used: N/A

🟡 Optional Information

Regression:
N/A — this is the first time this specific constraint has been checked, not a regression from a previously-working state.

Logs:
N/A — no error output; this is a constraint-consistency observation, not a failure report.

Additional Context:
Happy to send a one-line PR adding <2 to the eval extra's google-cloud-aiplatform bound if that's the intended fix, once confirmed.

How often has this issue occurred?:

• Always (100%) — the constraint as written is deterministic, not intermittent.