refactor(seahorse): use strings.Builder in compaction string helpers

[corporatepiyush]

📝 Description

The seahorse compaction helpers assemble prompt text from every message in a chunk using result += fmt.Sprintf(...) / content += s. That pattern is O(n²): each += allocates a new backing array and copies the whole accumulated string. This PR switches the four helpers to strings.Builder (amortised O(n), single growing buffer).

Output is byte-identical to the previous implementation (verified against the old code for n = 0, 1, 7, 200).

Functions changed (all unexported helpers, no API change):
formatMessagesForSummary, formatSummariesForCondensation, truncateSummary, truncateCondensedSummaries.

🗣️ Type of Change

• 🐞 Bug fix
• ✨ New feature
• 📖 Documentation update
• ⚡ Code refactoring (no functional changes, no api changes)

🤖 AI Code Generation

• 🤖 Fully AI-generated
• 🛠️ Mostly AI-generated (AI draft, Human verified/modified)
• 👨‍💻 Mostly Human-written

🔗 Related Issue

N/A — self-contained allocation cleanup on the context-compaction path.

📚 Technical Context

• Reasoning: += on a growing string is quadratic in the number of iterations; strings.Builder grows one backing buffer and appends in place. These helpers run during context compaction on chunks that can hold hundreds of messages, so this is directly on-brand for PicoClaw's low-memory goals.

Benchmarks (200 messages, go1.26, darwin/arm64, -benchmem):

function	before	after	speedup	memory
formatMessagesForSummary	375µs · 1.78 MB · 1004 allocs	53µs · 95 KB · 615 allocs	7.1×	−94.7%
truncateSummary	254µs · 1.24 MB · 204 allocs	10µs · 49 KB · 15 allocs	25×	−96%

Reproduce:

go test -run='^$' -bench='BenchmarkFormatMessagesForSummary|BenchmarkTruncateSummary' -benchmem ./pkg/seahorse/

(Both benchmarks are added in this PR.)

🧪 Test Environment

• Hardware: Apple Silicon (arm64)
• OS: macOS 15
• Model/Provider: N/A — pure refactor, no LLM calls exercised
• Channels: N/A

Verification: go vet ./pkg/seahorse/ and go test ./pkg/seahorse/ pass; gofmt -s clean. Existing parts_roundtrip_test.go (which asserts the output of formatMessagesForSummary and truncateSummary) still passes.

☑️ Checklist

• My code follows the style of this project.
• I have performed a self-review of my own changes.
• I have updated the documentation accordingly. (N/A — internal helpers, no documented behavior changes.)

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[github-actions]

This PR has had no activity for 7 days and has been marked as stale. If you are still working on it, please push an update or leave a comment; otherwise it will be closed automatically in 7 days.

[github-actions]

This PR has been closed after 14 days of inactivity. If you would like to continue, feel free to reopen it or submit a new PR.