feat: Add --insert flag to enable in-place TOC insertion

[ryo246912]

Fix: #26

This PR implements the --insert flag feature that allows users to insert generated TOC directly into markdown files, similar to the bash version of github-markdown-toc and doctoc.

Supported Commands

# Insert TOC into a file (creates backup by default)
gh-md-toc --insert README.md

# Insert TOC without creating backup
gh-md-toc --insert --no-backup README.md

[ekalinin]

Thanks for the PR @ryo246912, this is a nicely structured feature.
A few things I'd like to see addressed before merge.

Must fix

• Tests. There are no *_test.go files in the PR, and the two new adapters (filebackup.go, tocinserter.go) hold the most bug-prone logic (marker validation, replaceContent, insertAtTop). CONTRIBUTING.md sets adapters at >95% coverage, so could you add table-driven tests using t.TempDir()? Cases worth covering: markers present / absent / mismatched / inline, header and footer on/off, first-heading vs no-heading, and backup content + permissions.
• Non-atomic in-place write. InsertToc uses os.WriteFile, which truncates then writes. If the process is interrupted mid-write with --no-backup, the original file is lost with no recovery. Please write to a temp file in the same directory and os.Rename over the original (atomic on POSIX), keeping the backup as a secondary safeguard.

Should fix

• Marker detection vs replacement mismatch. hasMarkers / validateMarkers match with strings.Contains / strings.Count, but replaceContent only matches whole lines (trimmed == MarkerStart). A file with inline markers such as <!--ts--> old <!--te--> passes validation, then falls through replaceContent and returns a misleading "markers not found" error, failing the whole run. Please match consistently, or validate that markers must be on their own line.
• --insert with STDIN. STDIN is written to a temp file whose type is TypeLocalMD, so the insert path runs on it: it edits a file that is deleted right after, and CreateBackup leaves an orphaned ghtocXXXX.<timestamp> in the temp dir that the defer os.Remove does not clean up. Insert should be skipped for STDIN, the same way it is already skipped for non-local URLs in file.go.

Minor / nits

• validateMarkers: the if startCount == 0 && endCount == 0 { return nil } branch is dead code, since the function returns nil on the next line anyway.
• filebackup.go: os.Create uses default permissions instead of copying the source mode, and the second-resolution timestamp means two runs within the same second overwrite the earlier backup. InsertToc already preserves stat.Mode().Perm(), so it would be nice to be consistent.
• generateTimestamp embeds the username and current time, so re-running rewrites the file (and makes a new backup) even when the TOC is unchanged. This mirrors the bash tool, so it may be intended, just flagging it.
• insertAtTop treats any line starting with # as a heading, so a # comment inside a leading fenced code block would be picked as the insertion point. Low impact.
• Toc.String() uses a pointer receiver while At uses a value receiver, and it duplicates the row-join logic already in CustomPrint. Minor consistency point.

Overall a solid contribution that fits the architecture well. The tests and the atomic write are the main blockers for me. Thanks again!