[fix] allow long threshold expr when binding many monitors

[orangeCatDeveloper]

Fixes #4171

Binding more than ~30 monitors to one alert rule fails with a validation error, because each bound monitor appends an equals(__instance__, "...") clause (~65 chars) to the threshold expr, and the field was capped at 2048 characters — 2048 / 65 ≈ 30 bindings.

The fix widens AlertDefine.expr from @Size(2048)/@Column(2048) to @Lob with a @Size(65535) validation cap — the same pattern the codebase already uses for NoticeTemplate.content. Migration V182 alters the column on all three vendors (mysql longtext with an INFORMATION_SCHEMA idempotency guard matching V172/V180, postgresql TEXT, h2 CLOB); widening is lossless for existing rows. The custom validation message was deliberately dropped so the default hibernate-validator message stays localized.

Backward compatible: previously valid input remains valid, the new 65535 cap still rejects runaway payloads, and expr is admin-authored config so no new trust boundary.

Verified end-to-end against a real backend (fresh H2, Successfully applied 7 migrations ... now at version v182):

before (2048 cap):
  POST expr=1501 chars (35 bindings)  -> {"code":0,"msg":"Add success"}
  POST expr=2296 chars (50 bindings)  -> {"code":1,"msg":"Size.alertDefine.expr:size must be between 0 and 2048"}
                                          (verbatim the error reported in the issue)

after (lob + 65535 cap):
  POST the same 2296-char expr        -> {"code":0,"msg":"Add success"}
  read back                           -> expr length 2296, stored intact
  POST 65536 chars                    -> {"code":1,"msg":"Size.alertDefine.expr:size must be between 0 and 65535"}