ClaudeSummarizer and OpenAISummarizer subclasses

[bobular]

The Anthropic dependency is currently hardcoded in the pom.xml but should go into the base-pom at some point, of course. See to-dos

Tested and currently deployed with Claude on https://bmaccallum.plasmodb.org/plasmo.bmaccallum/app/record/gene/PBANKA_1009500#category:transcriptomics at Bob's expense ($10 API costs currently paid up-front out of pocket; $0.50 currently used - please test with PBANKA genes only. It's a VEuPathDB organisation token however. Let me know if you want to join the VEuPathDB Anthropic console organisation.)

New env vars

• CLAUDE_API_KEY (Anthropic is rebranding to Claude, by the way).
• DOLLAR_COST_PER_1M_AI_INPUT_TOKENS=3
• DOLLAR_COST_PER_1M_AI_OUTPUT_TOKENS=15
• MAX_DAILY_AI_EXPRESSION_DOLLAR_COST=33
  (Previous OPENAI_-prefixed cost vars are deprecated but still functioning.)

This was 90% Claude coded but with careful hand-holding all the way. It struggled massively with the message API response handling, until I showed it some example code from the github repo, then it was fine. Context7 docs via MCP didn't really help.

TO DOs

• handle anthropic SDK version in maven base pom properly
• set up conifer model props properly for deployment (I just manually edited palm:/home/bmaccallum/bmaccallum.plasmodb.org/gus_home/config/PlasmoDB/model.prop for my dev site)
• set up institutional payment
• testing on qa on all organisms
• consider "extended thinking" mode - see example code