In @.github/workflows/time.yml at line 20, Update the GitHub Actions checkout
step to use the newer actions/checkout@v4 instead of actions/checkout@v3; locate
the workflow step that currently contains the string "uses: actions/checkout@v3"
and change it to "uses: actions/checkout@v4" to pick up Node.js 20 support and
performance improvements.

In @.github/workflows/time.yml at line 22, Replace the blocking entropy source
in the workflow run command: update the run step that currently invokes (time
(head --bytes=1000000000 /dev/random >/dev/null)) 2>&1 | tee output.txt to use
/dev/urandom instead of /dev/random so the head --bytes invocation is
non-blocking and yields consistent benchmark timing.

In @.github/workflows/time.yml at line 36, The YAML value for the key
alert-comment-cc-users is invalid because it's written as a comma-separated
string ('@ktrz','@henrikingo'); change it to a proper YAML sequence or a
multiline string supported by the action. Replace the current value with either
a YAML array (e.g., alert-comment-cc-users: ["@ktrz", "@henrikingo"]) or a
pipe-delimited block (alert-comment-cc-users: | followed by each user on its own
line) so actionlint will parse the workflow correctly.

In @.github/workflows/time.yml at line 50, The YAML entry for
alert-comment-cc-users is using invalid syntax ("'@ktrz','@henrikingo'"); update
the alert-comment-cc-users value to a proper YAML sequence such as
alert-comment-cc-users: ['@ktrz', '@henrikingo'] or a block list (- '@ktrz' / -
'@henrikingo') so the workflow parser accepts it; edit the
alert-comment-cc-users line in the time.yml workflow to use one of these valid
list formats.

In `@action.yml` at line 14, Update the description string for the action (the
description key in action.yml) to include the missing valid tool names "jmh" and
"benchmarkluau" so the full list of allowed tools reflects all supported
options; simply add "jmh" and "benchmarkluau" into the quoted comma-separated
list in the existing description value.

In `@src/extract.ts` around lines 696 - 705, The parseTimeOutput function is
incorrectly using tparts[1].substring(-1) which returns the whole string
(including the trailing 's') and causes parseFloat to yield NaN; update
parseTimeOutput to remove the trailing 's' using tparts[1].slice(0, -1) before
calling parseFloat (and ensure you still guard for undefined t and tparts[1] as
currently done).

In `@src/extract.ts` around lines 746 - 751, The three benchmark results all use
the literal name values ('real', 'user', 'sys'), losing per-command context;
update the results.push calls (the objects constructed in the results.push in
this snippet) to include the command identifier from the existing extra/name
variable in the benchmark name (e.g. combine the command name variable "name"
with the metric label so the name field becomes commandName + '-real' (and
similarly '-user' and '-sys')), and ensure you make the same change for the
other two pushes that produce 'user' and 'sys' entries.

In `@src/extract.ts` around lines 728 - 781, The parsing of time outputs in
extractTimeBenchmarkResult (which calls parseTimeOutput) fails for locales using
commas as decimal separators (e.g., "0m0,100s"); update parseTimeOutput to
normalize locale decimals by replacing a single comma in the fractional part
with a dot (and strip any thousands separators) before calling parseFloat,
ensuring lines like "0m0,100s" and "0m0.100s" both parse correctly; adjust
parseTimeOutput (referenced by extractTimeBenchmarkResult) to perform this
normalization and add tests for comma and dot decimals.