fix(seer): Make widget conditions readable for the Seer Explorer agent

[Mihir-Mavalankar]

• Widget query conditions use internal \uf00d Unicode markers around wildcard operators (e.g. Contains, StartsWith), which produce garbled strings like Containsqueue.task.taskworker when rendered to markdown for the Seer agent — causing the agent to make broken tool calls with literal operator text as search values.
• Add readableConditions() that replaces the 6 \uf00d-delimited operators with readable labels (e.g. contains, starts with) via simple string replacement. All other query structure — AND, OR, parentheses, free text, negation, comparison operators — passes through unchanged.
• Update widget query hints to describe what each widget type visualizes (timeseries chart, table, single number) and direct the agent to understand intent from the query config, instead of referencing internal API parameters (y_axes, group_by) that the natural-language tool doesn't use.
• Add 20 tests covering all 6 operator types, partial match safety, mixed operators, repeated operators, OR with parens, free text passthrough, literal "Contains" without markers, and a real-world dashboard widget query.

[gggritso]

Makes sense 👍 I'll leave it to you if you want to make the query conversion more robust or not. I think if you don't cover all the cases now it'll be harder to spot the bad cases later, but up to you