ble: Add firmware revision to device info service

See merge request !496

Indicate on screen that a reset will happen

See merge request !495

Add a boot animation to show off CTX

See merge request !494

When ctx is currently running in the dispatcher task to perform some
drawing calculations, the code here in the pmic task might preempt it.
This can lead to odd glitches and probably in the worst case a crash.
Because we know that no other code will continue running after we've hit
the shutdown codepath, we can just force-reinitialize ctx here.

Print "Reset!" to the screen before resetting so it is obvious what is
about to happen.  Currently, the screen brightness just suddenly jumps
up and card10 becomes unresponsive until the bootloader shows up.

feat(ble): Add a queue for indications and notifications

See merge request !491

Closes #237

See merge request !490

Closes #238

See merge request !489

This makes it possible to reliably stream data (like ECG samples) to another device

Closes #237

Fixes #238

hack(ci): Disable cert check for now

See merge request !492

Looks like the docker image is too old for the new LE certs

fix(sleep): Keep HIRC turned on when switching back to HIRC96

Closes #236

See merge request !488

fix(mp): Enable gatt client functionality again

See merge request !487

Update CHANGELOG

See merge request !485

Closes #232

See merge request !486

Fix epic_sleep() calls delaying load of new app

See merge request !484

When core 1 repeatedly calls `epic_sleep()`, this could delay an
epic_load() from core 0 because the reset interrupt can only be
delivered after such a call completed.

To work around this, add calls into `do_load()` to notify the dispatcher
task that it should return early from calls like `epic_sleep()`.  This
gets a bit tricky because the dispatcher task also uses
`ulTaskNotifyTake()` to wake up when new API calls arrive.  To make sure
we still break out of `epic_sleep()`, we need to call
`xTaskNotifyGive()` twice in the wait loop.

card10.cfg: Add config option for flashlight LED

See merge request !483

CTX Rendering

See merge request !473

"Cache" the local sleep value to not send repetitive sleep-out commands
to the display when the user repeatedly sets the backlight.

Adapted from previous implementation in !476.  To quote:

> This saves about 1 mA when the display is off.

Previously, such lines would be drawn as a fuzzy two-pixel wide line
instead of a sharp one-pixel one.  This is because the coordinate grid
is aligned _between_ pixels instead of on the pixel centers.  More
information in this cairo FAQ:

    https://www.cairographics.org/FAQ/#sharp_lines



For odd linewidths, add 0.5 to all coordinates to offset the line
perfectly onto the pixel grid.  This makes them look sharp, as with the
previous gfx code.

Suggested-by: Øyvind Kolås <pippin@gimp.org>

Not sure why I didn't do this right away, but using 1.95 makes the
circles look much better.  Using 2 leads to a small artifact on the
right side so we have to stay slightly below...

Remove the last remaining piece of gfx integration.  With this, the
card10 firmware is now fully rendering using ctx instead!

Reimplement `epic_disp_blit()` without relying on the gfx library.

Make the new ctx based implementation behave just like the old one.
This means _including_ the yend row and xend column.

This fixes the G-Watch and DigiClk not looking as they should.

Use ctx to display the panic message.  The implementation contains a few
safeguards to hopefully deal with any breakage that happended before the
panic.

With the PMIC task calling graphics code, it now needs a bit more stack
space.  Increase the allocation to ensure we won't get a stack overflow
here.