Calculation bugs need to be fixed

[konard]

• Port doublets to stable Rust, migrate deps to crates.io, update CI/CD linksplatform/doublets-rs#48 (comment) (Context window: repetition can be removed, code can be unified, 7 / 64K (0%) output tokens, 4 / 64K (0%) output tokens looks like a bug for Haiku sub-agents sessions, as we do with sub-agents calls, we can also output in the title for Claude Opus 4.6: the number of sessions like (3 session segments)).
• Fix JS CI/CD: npm upgrade fallbacks and publish error detection web-capture#55 (comment) (282 / 64K (0%) output tokens looks like a bug here, also for some reason there is no detalization on haiku and sonnet separate sessions, as we do with Sonnet and other sub-agent sessions, note that multiple sessions go with numbered list, and if there only single session we should output bullet marked list)
• feat: add Opus 4.7 support with corrected effort mapping #1621 (comment) (we need to double check why total cost calculation is different by last digit, but our difference field does not show that, yet difference is visible in actual calculations, we need to ensure we use decimal.js for all tokens and pricing calculations, to avoid floating point precision related issues; also for some reason there is no detalization on haiku separate sessions, as we do with Sonnet and other sub-agent sessions, we need to unify the logic, so it works the same for all models of subagents and in all cases).

Claude Opus 4.6:

- Context window: 101.9K / 1M (10%) input tokens, 26.7K / 128K (21%) output tokens

For main model we can now omit Context window:, as it does not add value for the reader on multi session mode, and will allow make output unified with how sub-agents do it. So we can deduplicate the code as much as possible.

We also need to double check that all our calculations are correct in all cases.

We need to collect data related about the issue to this repository, make sure we compile that data to ./docs/case-studies/issue-{id} folder, and use it to do deep case study analysis (also make sure to search online for additional facts and data), list of each and all requirements from the issue, and propose possible solutions and solution plans for each requirement (we should also check known existing components/libraries, that solve similar problem or can help in solutions).