PURPOSE:
            List the actual search queries that triggered the account's Google Ads, so you can
            review wasteful or off-target terms and decide which ones to add to a negative
            keywords list. Mirrors the "Search Terms" library page in the Metadata UI.

            WHEN TO USE:
            - Investigate which queries are costing money without producing clicks or conversions
            - Find queries to add to a negative keywords list (pair with add_keywords_to_negative_lists)
            - Inspect search terms tied to a specific experiment or matched keyword
            - Sort by cost / clicks / conversions to surface the worst offenders

            FILTERS (all optional):
            - searchTerm: substring match on the search term text
            - experiment: experiment name substring match
            - keyword: matched keyword substring
            - matchType: array subset of [PHRASE, BROAD, EXACT, NEAR_EXACT, NEAR_PHRASE]
            - startDate / endDate: ISO-8601 timestamps (e.g. 2026-06-01T00:00:00.000Z).
              The UI defaults to the last 30 days when both are omitted.

            PAGINATION & SORT:
            - page: 0-based (default 0)
            - size: rows per page (default 12, max 100)
            - sort: "<field>,<direction>". Sortable fields include searchTerm, matchType,
              experiment, addedExcluded, keyword, impressions, clicks, ctr, averageCpc,
              spent, conversions, costPerConversion. Direction is asc or desc.
              Default: "clicks,desc".

            RESPONSE FORMAT:
            Returns a paginated response:
            {
              "totalElements": 123,
              "totalPages": 5,
              "data": [
                {
                  "id": "bGVhZCBnZW5lcmF0aW9uIGNhbmFkYQ",
                  "searchTerm": "lead generation canada",
                  "experiment": "MD_Experiment",
                  "experimentId": "179908",
                  "matchType": "PHRASE",
                  "keyword": "lead generation",
                  "addedExcluded": "Added",
                  "impressions": 200,
                  "clicks": 11,
                  "ctr": 5.5,
                  "averageCpc": 2.3,
                  "spent": 25.30,
                  "conversions": 0,
                  "costPerConversion": null
                },
                ...
              ]
            }

            COMMON USE CASES:
            1. Top spenders with zero conversions in the last 30 days:
               list_search_terms(sort="spent,desc")
               then filter the response client-side for conversions == 0.

            2. Worst CTR among phrase matches:
               list_search_terms(matchType=["PHRASE"], sort="ctr,asc")

            3. Search terms tied to a specific experiment:
               list_search_terms(experiment="Q2 Brand Test")

            4. Date-bounded review:
               list_search_terms(
                 startDate="2026-05-01T00:00:00.000Z",
                 endDate="2026-05-31T23:59:59.999Z",
                 sort="spent,desc"
               )

            PARAMETERS:
            - searchTerm (string, optional): substring filter on the search term text.
            - experiment (string, optional): experiment name substring filter.
            - keyword (string, optional): matched keyword substring filter.
            - matchType (array of strings, optional): subset of
              [PHRASE, BROAD, EXACT, NEAR_EXACT, NEAR_PHRASE].
            - startDate (string, optional): ISO-8601 timestamp lower bound.
            - endDate (string, optional): ISO-8601 timestamp upper bound.
            - page (integer, optional): 0-based page index (default 0).
            - size (integer, optional): rows per page (default 12).
            - sort (string, optional): "<field>,<direction>" (default "clicks,desc").

            CHANNEL SCOPE:
            BOTH search channels are supported: pass channel=MICROSOFT_ADS for Microsoft
            Ads (Bing) and channel=GOOGLE_ADS for Google. The two channels hold SEPARATE
            reports, so never substitute one channel for the other to get "close enough"
            results.