If users want a strobe mode, they usually want it for defence purposes.
Thus it should be possible to enter it as fast as possible. If strobe is
enabled (first option if STROBE is compiled) the flashlight always
starts in a strobe mode. Then a single press of the power switch enters
the usual UI.

TIMSK0 is outside of lower 32 I/O registers and thus cannot be set via
sbi. This saves 4 bytes of flash.

Set a flag in the interrupt handler instead of checking the current tick
value.

Requires additional 14 bytes of flash...

Now we use registers in the function call area. But we don't use any
functions with many arguments, so this is not a problem and saves 56
bytes.

Costs 8 bytes of flash.

Check fast_presses and state only once. Allows more deterministic
transitions, but introduces a goto and increases flash size by 14 bytes.

Move common actions out of mode specific handler. Removes redundant
code, but might not provide the same level of flexibility as before and
might reduce readability.

Saves 2 bytes of flash.

Just to make sure. We have enough free RAM at the moment.

toggle_options() preserves the value of options if it returns. The
compiler did not optimize the intermediate variable because it doesn't
know that fact. This saves 4 bytes.

Also make some tiny improvements to the Makefile related to fuses.

This allows more flexibility in choosing the brightness of blinking,
currently used only to flicker at the current brightness for battery
warning.

Sadly due to additional calls to set low intensity for config mode this
saves only 2 bytes of flash memory, but with some other parameters it is
possible to save up to 6 bytes more.

Using ATOMIC_FORCEON saves 2 bytes compared the ATOMIC_RESTORESTATE
because it does not need to save the previous state.

Thus all necessary delays fit into an uint8_t instead of uint16_t which
saves 18 bytes in flash.

- Use dedicated delay_s() function
- Move ticks into register

Now everything fits exactly in 1024 byte!

This is non-optimized code and too big for flash.

As the datasheet tells us, fast PWM cannot reach zero duty cycle.

D'oh!

This saves some bytes because it is not necessary anymore to address
options in RAM.

10 kHz is too fast for the 7135 current regulator which requires at
least ~5-8 µs to provide full current. With 2.3 kHz the shortes pulse
(PWM level 5) is ~9 µs. This results in higher efficiency and should
still be fast enough to not show visible flicker. The downside is that
this might cause audible humming.