Keep a fractional spawn rate in the --html report file name - #3555

Problem

--html documents {u}, {r}, {t} as being converted to "number of users, spawn rate and run
time respectively", but process_html_filename() cast every substituted value with int() first:

options.html_file = options.html_file.replace(option_term, str(int(option_value)))

-r/--spawn-rate is type=gt_zero(float) (locust/argument_parser.py:492), so a fractional rate is
valid input and the runner honours it — only the report name truncated it. -r 0.05, -r 0.2 and
-r 0.5 all produced the same results_10_0_5.html, so consecutive runs overwrote each other's
report, and the name contradicted the flag that was passed.

Fixes #3554.

Change

locust/html.py — keep the integer spelling for whole numbers and spell fractional values as
passed:

-            options.html_file = options.html_file.replace(option_term, str(int(option_value)))
+            # -r is a float option, so a fractional spawn rate must survive the substitution.
+            replacement = str(int(option_value)) if float(option_value).is_integer() else str(option_value)
+            options.html_file = options.html_file.replace(option_term, replacement)

{u} and {t} are unchanged in practice: --users is type=int, and parse_timespan() returns
whole seconds and rejects fractional input (90.5, 0.5s both raise), so {r} is the only tag
whose option is genuinely a float. -r 5 still parses to 5.0 and still yields
some_name_11_5_1.html, which is what test_main.py::test_html_report_option already asserts.

locust/test/test_html_filename.py — two cases appended to the existing
TestProcessHtmlFilename: the per-value expectation (0.5 -> 0.5, 1.5 -> 1.5, 2.0 -> 2) and
the collision case that motivated the fix (three distinct rates must not collapse onto one path).

How it was checked

On master at 35c06fad, Python 3.11.14, Windows:

• New tests before the change: 2 failed, 4 passed
  (AssertionError: 1 != 3 for the collision case; results_10_0.html instead of
  results_10_0.5.html).

• pytest locust/test/test_html_filename.py -q after the change: 6 passed.

• Measured through the real parser (get_parser().parse_args(...)) into process_html_filename(),
  which is the same object locust/main.py:674 hands over:

  -r	parsed	master	this branch
  0.05	0.05	results_10_0_5.html	results_10_0.05_5.html
  0.2	0.2	results_10_0_5.html	results_10_0.2_5.html
  0.5	0.5	results_10_0_5.html	results_10_0.5_5.html
  1.5	1.5	results_10_1_5.html	results_10_1.5_5.html
  5	5.0	some_name_11_5_1.html	some_name_11_5_1.html

• ruff check locust/: all checks passed; ruff format --check on both files: already formatted;
  mypy locust/html.py: no issues; typos on both files: clean (ruff 0.16.8, mypy 1.14.1).

• pytest locust/test/test_main.py -q on this machine: 50 failed / 19 skipped both with and
  without the change — pre-existing Windows-only failures (TestProcess raises
  TypeError: bytes args is not allowed on Windows), not caused by this diff.